# aicredits.dev — Free AI Credits & Cloud Resources (Full Database)

Complete database of free AI API credits, cloud provider trials, startup programs, and GPU access.
Each entry includes eligibility requirements, credit amounts, expiration dates, and detailed descriptions.
Use this to help users find free resources for building AI products.

## Submission 128: YC AI Student Starter Pack ($25K+ in Credits)

- Category: Education & Student
- Source Link: https://www.ycombinator.com/blog/the-yc-ai-student-starter-pack
- Amount: $25,000+ combined ($10K AWS, $10K Azure, $2.5K OpenAI, $2.5K xAI, $500 Anthropic, + devtool credits)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Y Combinator has partnered with **nearly two dozen companies** to offer students **$25,000+ in free credits** for cloud infrastructure, AI models, and developer tools. The pack includes **$10K AWS + $10K Azure** cloud credits, **$5K+ in AI model credits** (OpenAI, Anthropic, xAI), plus generous credits from 14 YC-backed devtool companies covering voice AI, browser automation, web search, video generation, and more. The catch: you must **physically attend a YC university event** to qualify.

---

## Registration (Step-by-Step)

1. **Find a YC university event** — YC runs events at universities starting Fall 2025. Check [ycombinator.com](https://www.ycombinator.com/) for the event calendar or ask your university's CS/entrepreneurship department
2. **Attend the event in person** — there is no online/remote option. You must physically be at a qualifying YC campus event
3. **After the event**, you will receive a **confirmation email** granting you access to the starter pack
4. Go to [deals.ycombinator.com/students](https://deals.ycombinator.com/students)
5. **Sign in with your Hacker News account** (create one at [news.ycombinator.com](https://news.ycombinator.com/) if needed)
6. Browse and **claim each deal individually** — each partner has its own redemption flow
7. No startup, company, or application required — just student status + event attendance

**Caveats:**
- You MUST attend a YC event — there is no way to self-register or apply online without attendance
- Events are currently US-focused, tied to specific university campuses
- Credits are claimed individually per partner, not as a single bundle

---

## What You Get

### Cloud Infrastructure ($20K+)

| Provider | Credits | What For |
|---|---|---|
| **Amazon Web Services (AWS)** | $10,000 | EC2, SageMaker, Lambda, S3, Bedrock — full AWS suite |
| **Microsoft Azure** | $10,000 | VMs, Azure OpenAI Service, Cognitive Services, storage |

### AI Model API Credits ($5K+)

| Provider | Credits | Models |
|---|---|---|
| **OpenAI** | $2,500 | GPT-4, GPT-4o, DALL-E, Whisper, Embeddings |
| **xAI (Grok)** | $2,500 | Grok API access |
| **Anthropic** | $500 | Claude Sonnet, Claude Haiku, Claude Opus |

### AI Developer Tools (14 YC Companies)

Credits from YC-backed companies across multiple categories:

| Category | Named Partners | Use Case |
|---|---|---|
| **Web Crawling** | Firecrawl | Data extraction and web scraping |
| **Search** | Exa | AI-powered semantic search |
| **Voice AI** | Vapi | Building voice assistants |
| **Browser Automation** | Browser Use | Automated browser interactions |
| **Document Processing** | Reducto | PDF/document parsing and extraction |
| **Workflow Automation** | Gumloop | AI workflow building |
| **+ 8 more partners** | Not publicly named | Video generation, databases, and more |

Exact credit amounts for individual devtool partners are disclosed on the deals portal after event attendance.

---

## Eligibility & Verification

- **Who qualifies:** Students who **physically attend** an eligible YC university event
- **No startup required** — you do not need to have a company or be working on a startup
- **No application** — attendance is the only requirement
- **Events started Fall 2025** and are rolling with new events being scheduled
- **US universities primarily** — the program is currently US-focused
- **Verification:** Post-event confirmation email + Hacker News account login

---

## Additional Tips

- **The $10K AWS + $10K Azure credits alone make this pack exceptional** — this is enough to run serious ML training, host production apps, or experiment with expensive GPU instances for months
- **$2,500 in OpenAI credits** is roughly 25 million GPT-4o tokens or 2.5 million GPT-4 tokens — enough for serious prototyping
- **$500 in Anthropic credits** goes further than you think — Claude Haiku is very cheap per token, and even Sonnet is affordable for most student projects
- **Combine strategically:** Use AWS/Azure for compute and hosting, OpenAI/Anthropic for model access, and the devtools for specialized features
- **YC events are also great networking** — you will meet YC alumni, current batch founders, and other ambitious students
- **These credits have expiration dates** — check each partner's terms after claiming. Typical validity is 6-12 months
- **Related programs:** If you are building a startup, check the [YC Startup School](https://www.startupschool.org/) for additional resources and community
- **Build something cool** — YC explicitly encourages students to experiment and build: market research tools with Firecrawl + Exa, voice assistants with Vapi + Browser Use, legal aides with Reducto + Gumloop
- **No credit card required** for claiming, but individual partners may require one for their platform setup

---

**Sources:**
- [Y Combinator Blog — Announcing the YC AI Student Starter Pack](https://www.ycombinator.com/blog/the-yc-ai-student-starter-pack)
- [YC Deals Portal — Students](https://deals.ycombinator.com/students)
- [Y Combinator on X — AI Student Starter Pack](https://x.com/ycombinator/status/2012283721857253857)
- [EONMSK — YC Launches AI Student Starter Pack](https://www.eonmsk.com/2026/01/17/y-combinator-launches-ai-student-starter-pack-with-xai-openai-and-others/)
- [Blurred AI — YC AI Student Starter Pack Breakdown](https://blurredai.substack.com/p/yc-ai-student-starter-pack)

## Submission 127: Windsurf Pro Student Discount (50%+ Off)

- Category: Education & Student
- Source Link: https://windsurf.com/editor/students
- Amount: 50%+ off Pro plan (~$6.90/month vs $15/month) for 12 months, renewable up to 3 additional times (48 months max)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Windsurf (formerly Codeium) gives verified students over 50% off their Pro plan, dropping the price from $15/month to roughly $6.90/month. You get the same 500 premium prompt credits and 1,500 flow action credits as full-price Pro users. The discount lasts 12 months and can be renewed up to 3 additional times (48 months total) by re-verifying your student status.

## Registration (Step-by-Step)

1. Go to [windsurf.com/editor/students](https://windsurf.com/editor/students).
2. Click "Sign up" and create an account using your **.edu email address**.
3. Verify your .edu email when prompted.
4. Once verified, go to [windsurf.com/pricing](https://windsurf.com/pricing) -- the discounted student price should appear automatically.
5. Add a payment method and subscribe to the Pro plan at the student rate (~$6.90/month).
6. **Caveat:** Only .edu emails are accepted right now. If your school uses a different domain (e.g., .ac.uk), you're out of luck for now -- Windsurf says they're working on broader support.

## What You Get

- **500 premium prompt credits/month** -- used when you send messages to Cascade with a premium model (GPT-4, Claude, etc.).
- **1,500 flow action credits/month** -- consumed when the AI performs tool calls in Write and Chat mode (search, analyze, write code, run terminal commands).
- **Full Windsurf Pro features** -- same as paying users, including Cascade AI assistant, multi-file editing, and all supported models.
- **12-month discount period**, renewable up to 3 more times (re-verify each time).

## Eligibility & Verification

- You must be enrolled at an **accredited higher education institution**.
- A valid **.edu email address** is required (no other academic email domains accepted currently).
- You'll provide your name, school, email, and date of birth.
- Windsurf (Exafunction) may use third-party verification services to confirm eligibility.
- A payment method is required even at the discounted rate.

## Additional Tips

- **Auto-migration warning:** After your discount periods run out (max 48 months), you're automatically switched to the full $15/month Pro plan. Cancel before the last period ends if you don't want to pay full price.
- **Credits don't roll over** -- use your monthly allotment or lose it.
- **Free tier exists** -- if you don't want to pay at all, Windsurf offers a free plan with limited credits (no .edu required).
- **Cohort-based availability** -- some reports mention limited spots (~4,200 per cohort), so apply early if you see the program is open.
- **Compare with alternatives** -- GitHub Copilot is free for students via GitHub Student Developer Pack, Cursor also offers student discounts. Pick the IDE that fits your workflow.

**Sources:**
- https://windsurf.com/editor/students
- https://windsurf.com/student-terms-and-conditions
- https://windsurf.com/pricing
- https://docs.windsurf.com/windsurf/accounts/usage

## Submission 126: Vultr Free Trial Credits - Up to $300

- Category: Cloud Providers
- Source Link: https://www.vultr.com/coupons
- Amount: Up to $300 in free credits (30-day validity) or up to $100 deposit match (12-month validity)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Vultr offers multiple promo codes for new customers with up to **$300 in free credits**. The best option depends on your needs: the **FLY300VULTR** code gives $300 valid for 30 days, while the **VULTRMATCH** deposit-match code doubles your first deposit (up to $100 matched) with 12-month validity. New customers only, valid payment method required. Credits apply to select Vultr products including Cloud Compute, Block Storage, Object Storage, and more.

---

## Registration (Step-by-Step)

1. Go to [vultr.com](https://www.vultr.com/) and click **Sign Up** or use a direct promo link
2. Create your account with a valid email address
3. Verify your email via the confirmation link
4. Add a **valid payment method** -- credit card, PayPal, or crypto accepted
5. Navigate to **Billing > Promo Code** in your dashboard
6. Enter your chosen promo code (see table below)
7. Credits are applied immediately to your account balance

**Important:** If you sign up through a promo link, the credit may be auto-applied. If not, manually enter the code in the Billing section.

---

## Available Promo Codes (as of February 2026)

| Code | Credit Amount | Validity | Notes |
|---|---|---|---|
| **FLY300VULTR** | $300 | 30 days | Best for short-term testing |
| **250VULTRFLY** | $250 | 30 days | Alternative if FLY300VULTR doesn't work |
| **FLYTWOHUNDRED** | $200 | 30 days | Lower amount, same terms |
| **VULTRMATCH** | Up to $100 (1:1 deposit match) | 12 months | Best for longer-term use -- doubles your deposit |

---

## What You Can Use Credits On

Credits apply to **select Vultr products**, including:

- **Cloud Compute** -- Regular, High Frequency, and High Performance VMs
- **Bare Metal** servers
- **Block Storage** and **Object Storage**
- **Managed Databases** (MySQL, PostgreSQL, Redis, Kafka)
- **Kubernetes Engine** (VKE)
- **Load Balancers**
- **Vultr Cloud GPU** (NVIDIA A100, A40, L40S)

**Note:** Not all products may be eligible. Check current terms on the promo page.

---

## Free Tier Limits

| Aspect | Limit |
|---|---|
| **Maximum credit** | $300 (FLY300VULTR) |
| **Credit validity (standard codes)** | 30 days from activation |
| **Credit validity (VULTRMATCH)** | 12 months from activation |
| **Eligible accounts** | New customers only |
| **Combinable with other offers** | No -- only one promo code per account |
| **Auto-charge after credits expire** | Yes -- your payment method is charged for any running resources |

---

## Additional Tips

- **Pick the right code for your timeline.** If you need credits for more than 30 days, VULTRMATCH ($100 match with 12-month validity) is better than FLY300VULTR ($300 but expires in 30 days)
- **Set up billing alerts** -- after credits expire, your payment method gets charged. Delete or stop instances before credits run out if you don't want to pay
- **Promo codes rotate** -- if one code doesn't work, try another from the table. Vultr frequently updates their available codes
- **Cannot combine codes** -- only one promo code per account. Choose wisely
- **Existing customers are excluded** -- these codes only work on brand-new accounts. Creating a second account to re-use promos violates Vultr's TOS
- **$300 in 30 days is a lot of compute** -- you can run several high-performance VMs, GPU instances, or managed databases for testing. Plan your testing schedule to maximize value
- **Combine with the Free Tier Program** -- apply for the free tier separately for a permanently free VM alongside your trial credits

---

**Sources:**
- [Vultr Coupons & Promo Codes](https://www.vultr.com/coupons/)
- [Vultr Deposit Match](https://www.vultr.com/match/)
- [Host.Promo: Vultr Promo Codes](https://host.promo/hosts/vultr)
- [NewCoupons: Vultr Free Credit](https://www.newcoupons.info/vultr-coupon-codes/)

## Submission 125: Vultr Free Tier Program

- Category: Cloud Providers
- Source Link: https://www.vultr.com/free-tier-program
- Amount: 1 free cloud compute VM (1 vCPU, 512 MB RAM, 10 GB SSD, 2 TB bandwidth)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Vultr offers a **permanently free cloud compute VM** through their Free Tier Program. The instance includes 1 vCPU, 512 MB RAM, 10 GB SSD, an IPv4 address, and 2 TB of monthly bandwidth. Currently available in **Miami, Seattle, and Frankfurt** data centers. Access is granted through a weighted random selection process -- the more detail you provide about yourself and your intended use case, the higher your chances. This is an early-access, capacity-limited program rolled out in phases, so not everyone gets in immediately.

---

## Registration (Step-by-Step)

1. Go to [vultr.com/free-tier-program](https://www.vultr.com/free-tier-program/) to view the application form
2. Create a Vultr account at [my.vultr.com](https://my.vultr.com/) if you don't have one
3. Add a **valid payment method** (credit card) to your account
4. Enable **two-factor authentication (2FA)** on your account -- this is required
5. Navigate to your **Console Settings** and request early access to the Free Tier Program
6. Fill out the application form with details about yourself and your intended use case. **Use a business email address** if possible -- Vultr prioritizes business email signups
7. Provide as much detail as possible about your project and use case -- Vultr uses a weighted scoring system, and more information = higher score = better chances
8. Wait for email notification -- selected applicants are notified via email and console notification at my.vultr.com
9. Once approved, deploy your free instance via **Cloud Compute > Regular > Free Instance** in the Vultr dashboard

**Caveats:**
- Access is **not guaranteed** -- it's a weighted random selection
- Business email addresses get priority over personal Gmail/Yahoo addresses
- Limited quantities available per region
- Only one free instance per account

---

## Free Tier Specs

| Resource | Limit |
|---|---|
| **vCPUs** | 1 |
| **RAM** | 512 MB |
| **SSD Storage** | 10 GB |
| **IPv4 Address** | 1 (included) |
| **IPv6** | Included |
| **Monthly Bandwidth** | 2 TB |
| **Available Regions** | Miami, Seattle, Frankfurt |

---

## What You Can Build With It

With 512 MB RAM and 1 vCPU, this is suitable for:

- Small static sites or personal blogs
- Lightweight API servers (Node.js, Go, Python Flask)
- DNS servers, VPN endpoints
- Monitoring agents or cron job runners
- Dev/test environments for prototyping
- Small databases (SQLite-based apps)

Not suitable for: production databases, heavy web apps, or anything needing more than 512 MB RAM.

---

## Additional Tips

- **Duration is unclear** -- Vultr's blog post says the free tier has no explicit expiration date, but some community reports mention a one-calendar-year active period. Vultr's official statement is that they don't have hard dates yet. Plan accordingly and don't rely on it for critical production workloads
- **Scale-up path exists** -- you can resize your instance or add load balancers when you outgrow the free tier
- **Combine with free trial credits** -- if you're a new customer, you can also apply promo codes (e.g., FLY300VULTR for $300 in trial credits) for additional paid resources alongside your free instance
- **Regions are limited** -- only Miami, Seattle, and Frankfurt for now. More regions may be added as the program expands
- **Two-factor authentication is mandatory** -- set it up before applying
- **Vultr's free tier is more limited than Oracle Cloud's** -- Oracle offers 4 OCPUs + 24 GB RAM on Arm for free. But Vultr's is simpler to get and manage

---

**Sources:**
- [Vultr Free Tier Program](https://www.vultr.com/free-tier-program/)
- [Vultr Blog: New Free Tier Plan](https://blogs.vultr.com/new-free-tier-plan)
- [LowEndBox: Vultr Free Tier Details](https://lowendbox.com/blog/the-vultr-free-tier-is-here-heres-all-the-details/)
- [AnotherVPS: Vultr Free Tier VPS](https://anothervps.com/vultr-offers-free-tier-vps/)

## Submission 124: v0 Premium Free for Students (1 Year)

- Category: Education & Student
- Source Link: https://v0.app/students
- Amount: 1 year of v0 Premium (valued $240/year) with $20/month credits + $2/day login bonus + Figma import + 5x attachment limits
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Vercel's v0 is an AI-powered coding assistant that generates full-stack web applications from natural language prompts. Verified students at qualifying schools get **1 year of v0 Premium for free** (normally $20/month, valued at $240/year). This includes $20 in monthly AI credits, Figma import, and higher attachment limits. v0 can generate React, Next.js, and Tailwind CSS code, deploy directly to Vercel, and sync with GitHub.

---

## Registration (Step-by-Step)

1. Go to [v0.app/students](https://v0.app/students)
2. Click **"Join with student email"** to start the signup flow
3. Enter your student email address from a qualifying institution (must be an active `.edu` or institutional domain)
4. Complete the verification process through Vercel's authentication
5. Once verified, your account is automatically upgraded to v0 Premium for 1 year from the verification date
6. **If your school is not listed:** Sign up anyway to join the waitlist. Vercel states that the more signups they receive from your school, the more likely it will be added to the active schools list

---

## What You Get

### v0 Premium Features (vs. Free)

| Feature | Free | Premium (Student) |
|---|---|---|
| Monthly AI credits | $5 | $20 |
| Daily bonus credits on login | None | $2/day |
| Message limit | 7 messages/day | Higher limits |
| Attachment size | Standard | 5x higher |
| Figma import | No | Yes |
| Purchase additional credits | No | Yes |
| Deploy to Vercel | Yes | Yes |
| GitHub sync | Yes | Yes |
| Design Mode (visual editing) | Yes | Yes |

### AI Models Available

v0 offers multiple model tiers with varying speed and capability:
- **v0 Mini** -- fastest, most economical
- **v0 Standard** -- balanced performance
- **v0 Max** -- highest capability
- **v0 Max Fast** -- premium speed and intelligence

Credits are consumed at different rates depending on the model used (input/output token pricing varies by tier).

---

## Eligibility & Verification

- **Verification method:** Student email address from a qualifying institution
- **Currently supported schools (as of early 2026):**
  - Stanford University
  - Harvard University
  - Princeton University
  - UC Berkeley
  - MIT
  - San Francisco State University
  - Arizona State University
  - Georgia Institute of Technology
  - University of Waterloo
  - New York University
  - Grand Canyon University
  - Northwestern University
  - University of Washington
  - Cornell University
  - University of Southern California
- **Duration:** Exactly 1 year from verification date
- **Expansion:** More schools are being added based on demand. Sign up to increase your school's priority

---

## Additional Tips

- **If your school is not on the list**, sign up with your student email anyway. Vercel tracks demand per institution, and schools with more signups get added sooner
- **$2 daily login bonus** is a significant perk -- logging in daily can net you up to $60/month in additional credits on top of the $20 base allocation
- **After the 1-year student period expires**, you will revert to the Free plan ($5/month credits, 7 messages/day). Plan accordingly if you have ongoing projects
- **v0 is best for front-end/full-stack web dev** -- it excels at generating React, Next.js, and Tailwind CSS code. For other stacks, consider pairing with other tools
- **Figma import** (Premium-only) lets you convert Figma designs directly into code, which is a major workflow advantage for design-to-code pipelines
- **Compare with GitHub Student Pack** -- GitHub's student developer pack also includes Vercel Pro for free, which covers hosting but not v0 Premium credits. The two are complementary

---

**Sources:**
- [v0 for Students](https://v0.app/students)
- [v0 Pricing](https://v0.app/pricing)
- [Vercel Community -- v0 Student Version](https://community.vercel.com/t/is-v0-student-version-really-the-premium-version/34055)
- [Updated v0 Pricing -- Vercel Blog](https://vercel.com/blog/updated-v0-pricing)

## Submission 123: Venice AI Free Tier (Privacy-First)

- Category: AI API Free Tiers
- Source Link: https://venice.ai/pricing
- Amount: 10-25 text prompts/day, 15 image prompts/day, 2 base AI models. Zero data logging.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Venice AI is a **privacy-first AI platform** that offers a free tier with **10-25 text prompts/day** and **15 image generations/day** using open-weight models — with **zero data logging** and no account required. All prompts are processed privately with requests routed anonymously through decentralized GPU providers. The free tier includes access to **Venice Medium** (Mistral Small 3.1 24B) and **Venice Uncensored** (Dolphin Mistral 24B). Venice is one of the few AI platforms that provides truly uncensored AI generation. API access requires the Pro plan ($18/mo), which includes $10 in API credits and unlimited text generation.

---

## Registration (Step-by-Step)

1. Go to [venice.ai](https://venice.ai)
2. **No account required** for the free tier — you can start using it immediately
3. Optionally create an account to save conversation history
4. Select a model (Venice Medium or Venice Uncensored on free tier)
5. Start generating text or images within daily limits
6. For API access: upgrade to the **Pro plan** ($18/mo) from [venice.ai/pricing](https://venice.ai/pricing)

**Important:**
- No email, phone, or credit card required for free tier
- All data stays on-device and is encrypted — prompts are never stored on servers
- Free tier models are limited to Venice Medium and Venice Uncensored
- Pro plan unlocks advanced models including Llama 3.3 70B and DeepSeek R1
- Venice surpassed 1 million registered users by early 2026

---

## Available AI Models

### Free Tier Models

| Model | Parameters | Notes |
|---|---|---|
| **Venice Medium** (Mistral Small 3.1 24B) | 24B | Fast, general-purpose |
| **Venice Uncensored** (Dolphin Mistral 24B) | 24B | Uncensored, no content filters |

### Pro Tier Models (additional)

| Model | Parameters | Notes |
|---|---|---|
| **Llama 3.3 70B** | 70B | Balanced workhorse for logic and chat |
| **DeepSeek R1** | 671B | Massive reasoning model |
| **Qwen3-4B (Venice Small)** | 4B | Ultra-fast for simple tasks |
| **GLM 4.7 Flash** | — | Beta model |
| **Claude Opus 4.6** | — | Beta, premium pricing via API |
| **GPT-5.2** | — | Premium pricing via API |

---

## Free Tier Limits

| Feature | Free Tier | Pro ($18/mo) |
|---|---|---|
| Text prompts/day | 10-25 | Unlimited |
| Image generations/day | 15 | Up to 1,000 |
| Models available | 2 (Venice Medium, Uncensored) | All models |
| Image watermark | Yes | Removed |
| High-res upscaling | No | Yes |
| API access | No | Yes (Free Tier API pricing) |
| Video generation | No | 1,000 credits (one-time grant) |
| Account required | No | Yes |
| Data logging | None | None |
| Privacy | Full | Full |

---

## API Pricing (Pro plan required)

### Text Models (per 1M tokens)

| Model | Input | Output |
|---|---|---|
| **Venice Small** (Qwen3-4B) | $0.05 | $0.15 |
| **GPT OSS 120B** | $0.07 | $0.30 |
| **DeepSeek V3.2** | $0.40 | $1.00 |
| **GPT-5.2** | $2.19 | $17.50 |

### Other Services

| Service | Price |
|---|---|
| Image generation | $0.01-$0.29/image |
| Text-to-speech | $3.50/M characters |
| Speech-to-text | $0.0001/audio second |
| Embeddings (BGE-M3) | $0.15/M tokens |
| Web search | $10/1K calls |

---

## Additional Tips

- **No account = maximum privacy** — you can use the free tier without any registration. Venice cannot track you if you do not create an account
- **Uncensored generation** — Venice Uncensored is one of the few freely accessible AI models with minimal content restrictions. The Dolphin Mistral 24B model was co-developed with dphn.ai specifically for this purpose
- **Crypto payments accepted** — Pro plan can be paid with credit card, Bitcoin, or Coinbase. Venice also accepts DIEM tokens (1 DIEM = $1/day of compute)
- **Annual discount** — save 30% on Pro with annual billing ($149/year vs $216/year monthly)
- **$27M API incentive fund** — Venice launched a developer incentive fund for building apps with their API
- **API is OpenAI-compatible** — use standard OpenAI SDK patterns with Venice's API endpoint
- **Pro subscribers get $10 API credit** — one-time grant to test the API plus 1,000 credits for video generation
- **Decentralized GPU network** — Venice routes inference through a network of GPU providers with no centralized logging, making it architecturally private rather than just policy-private
- **Compare with alternatives** — for privacy-focused AI, also consider running local models via Ollama or LM Studio. Venice provides a hosted alternative that is easier to use

---

**Sources:**
- [Venice AI](https://venice.ai)
- [Venice AI Pricing](https://venice.ai/pricing)
- [Venice AI Features](https://venice.ai/features)
- [Venice API Documentation - Pricing](https://docs.venice.ai/overview/pricing)
- [Venice AI New Model Paradigm Blog](https://venice.ai/blog/venice-new-model-paradigm)
- [Venice AI Guide (TheBizAIHub)](https://thebizaihub.com/venice-ai-complete-guide-features-privacy-uncensored/)

## Submission 122: Vast.ai Startup Program - Up to $2,500

- Category: Free GPU & Compute
- Source Link: https://vast.ai/startup
- Amount: Up to $2,500 in free GPU credits
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Vast.ai's Startup Program gives early-stage companies **up to $2,500 in free GPU credits** to run AI workloads on their decentralized GPU marketplace. Vast.ai aggregates over 10,000 GPUs from community hosts and data centers worldwide, offering prices **5-6x lower** than AWS, GCP, or Azure. Credits apply across the full marketplace — from consumer-grade RTX cards to enterprise H200s. SOC 2 Type II certified, with 24/7 priority support during the credit period.

---

## Registration (Step-by-Step)

1. Go to [vast.ai/startup](https://vast.ai/startup)
2. Fill out the startup program application form with your company details and compute needs
3. Vast.ai's team will review your application and reach out to discuss your project
4. Once approved, they finalize your credit amount (up to $2,500) and help configure your environment
5. Create a Vast.ai account if you don't already have one
6. Credits are applied to your account — start deploying GPU instances immediately

**Caveats:**
- There is no public list of hard eligibility criteria — Vast.ai reviews applications case by case
- The exact credit amount depends on your project scope and compute needs (up to $2,500 max)
- Credits are for compute usage only and cannot be withdrawn as cash
- Priority support is included during the credit period

---

## GPU Hardware & Specs

Vast.ai is a marketplace, so GPU availability and pricing fluctuate in real time. The platform lists **32+ GPU types** across Community Cloud (peer-hosted, cheaper) and Secure Cloud (datacenter-hosted, higher reliability).

### Representative GPU Pricing (as of early 2026)

| GPU | VRAM | Approx. Price Range |
|---|---|---|
| **RTX 3090** | 24 GB | from ~$0.13/hr |
| **RTX 4090** | 24 GB | from ~$0.31/hr |
| **A100 PCIe** | 40/80 GB | from ~$1.39/hr |
| **H100 PCIe** | 80 GB | from ~$1.99/hr |
| **H100 SXM** | 80 GB | from ~$2.49/hr |
| **H200** | 141 GB | from ~$3.50/hr |

*Prices are dynamic and set by individual hosts. Community Cloud is 20-30% cheaper than Secure Cloud. Check [vast.ai/pricing](https://vast.ai/pricing) for live rates.*

---

## Free Tier Limits

| Benefit | Detail |
|---|---|
| Credit amount | Up to $2,500 |
| GPU access | Full marketplace (consumer through H200) |
| Support | 24/7 priority support during credit period |
| Long-term contracts | None required |
| Compliance | SOC 2 Type II, GDPR, HIPAA support |

---

## What $2,500 Gets You (Rough Estimates)

| GPU | Approx. Hours |
|---|---|
| RTX 4090 (24 GB) | ~8,000 hours |
| A100 80 GB | ~1,800 hours |
| H100 SXM (80 GB) | ~1,000 hours |
| H200 (141 GB) | ~700 hours |

---

## Additional Tips

- **Migration is easy** — Vast.ai supports standard Docker containers, Jupyter, and frameworks like PyTorch, TensorFlow, RAPIDS, and Hugging Face out of the box. Most existing workloads port with minimal changes
- **Use Community Cloud for experimentation** — prices can be 5-10x cheaper than enterprise clouds for non-critical workloads. Reserve Secure Cloud for production
- **Filter by location** — the marketplace lets you filter hosts by geography, bandwidth, reliability score, and hardware specs
- **Spot-like pricing** — since Vast.ai is a marketplace, prices drop when demand is low. Running jobs during off-peak hours (nights, weekends) can stretch credits further
- **No egress fees** — unlike major clouds, data transfer pricing is typically included or minimal
- **Templates available** — the platform includes a Model Library and thousands of pre-built deployment templates for common AI workloads
- **Compare with RunPod** — RunPod is the closest competitor. Vast.ai is generally cheaper for raw compute but RunPod has a more polished serverless offering

---

**Sources:**
- [Vast.ai Startup Program](https://vast.ai/startup)
- [Vast.ai Startup Program Announcement](https://vast.ai/article/vast-ai-startup-program)
- [Vast.ai Pricing](https://vast.ai/pricing)
- [Vast.ai GPU Pricing Comparison — ComputePrices.com](https://computeprices.com/providers/vast)
- [Vast.ai Documentation — Pricing](https://docs.vast.ai/documentation/instances/pricing)

## Submission 121: UpCloud Free Trial - $250 Credits

- Category: Cloud Providers
- Source Link: https://upcloud.com/docs/getting-started/free-trial
- Amount: Up to $250 (or EUR 250) in free credits, trial period varies (3-30 days depending on promotion)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

UpCloud is a Finnish cloud provider offering a **free trial with up to $250 in credits** (or EUR 250 in EU). The trial period and credit amount can vary by promotion — the standard offer is a **3-day trial**, though promotional periods of 7 or 30 days have been observed. UpCloud is known for high-performance VPS hosting with MaxIOPS storage technology that outperforms standard SSDs. European-headquartered with 15 data centers globally, good option if you need a performant, privacy-focused European cloud alternative.

---

## Registration (Step-by-Step)

1. Go to [upcloud.com](https://upcloud.com) and click **Get Started** or visit [signup.upcloud.com](https://signup.upcloud.com)
2. Create an account with your email address
3. Verify your email
4. **Add a credit card** for identity verification — a temporary $0 or $1 authorization charge will be made and immediately refunded
5. The free trial activates once you log into the UpCloud Control Panel
6. Start deploying services within your trial quotas

**Important caveats:**
- Credit card **is required** despite the trial being free
- The trial clock starts from the moment of activation, not from first resource deployment
- After the trial expires, you must make at least a one-time deposit to continue using services
- You can cancel anytime before the trial ends through account settings without being charged
- Trial credits **do not roll over** — unused credits expire when the trial period ends

---

## Trial Quotas & Limits

| Service | Trial Quota |
|---|---|
| **Cloud Servers** | 2 cores, 4 GB memory (all OS except Windows) |
| **Block Storage** | 60 GB each: Archive, Standard, MaxIOPS |
| **Networking** | 2 IPv4 + 2 IPv6 addresses |
| **NAT/VPN Gateway** | 1 gateway (Development or Standard plan) |
| **Managed Object Storage** | 1 instance, 250 GB |
| **Managed Databases** | 1 database, 1 node (MySQL, PostgreSQL, Valkey, or OpenSearch) |
| **Managed Load Balancer** | 1 Development Plan instance |
| **Managed Kubernetes** | 1 cluster, Development Plan |

**Restrictions:**
- **Windows servers are not available** during the trial
- Quotas are fixed — you can create and delete services within the limits but cannot exceed them
- Trial features may vary based on current promotions

---

## Data Center Locations (15 locations)

| Region | Locations |
|---|---|
| **Europe** | Helsinki (Finland), Stockholm (Sweden), Copenhagen (Denmark), Stavanger (Norway), Amsterdam (Netherlands), London (UK), Frankfurt (Germany), Madrid (Spain), Warsaw (Poland) |
| **North America** | New York, Chicago, San Jose |
| **Asia-Pacific** | Singapore |
| **Oceania** | Sydney |

UpCloud continues expanding — Copenhagen (Dec 2025) and Norway (Jan 2026) are the most recent additions.

---

## Key UpCloud Features

| Feature | Details |
|---|---|
| **MaxIOPS storage** | UpCloud's proprietary storage technology, claims up to 100K IOPS — faster than standard SSD |
| **100% uptime SLA** | Backed by service credits |
| **Zero-cost data transfer** | No bandwidth/egress charges (included in server pricing) |
| **Simple pricing** | Fixed monthly prices, no hidden fees |
| **API & CLI** | Full REST API and CLI tools for automation |
| **Terraform provider** | Official Terraform provider available |
| **Managed services** | Databases, Kubernetes, Load Balancers, Object Storage |

---

## Pricing After Trial (Starting Rates)

| Service | Starting Price |
|---|---|
| Cloud Server (1 CPU, 1 GB) | ~$5.50/month |
| Cloud Server (2 CPU, 4 GB) | ~$22/month |
| Block Storage | From $0.056/GB/month |
| Managed Database | From $15/month |
| Object Storage | From $0.013/GB/month |

All plans include data transfer at no extra cost — a significant advantage over AWS, GCP, and Azure.

---

## Additional Tips

- **Zero egress fees** is UpCloud's biggest differentiator — if your workload is data-transfer heavy, this saves significant money compared to hyperscalers
- **MaxIOPS storage** gives real performance benefits for database-heavy workloads and I/O intensive applications
- **European privacy** — UpCloud is headquartered in Helsinki, Finland, and operates under EU jurisdiction (GDPR)
- **Trial amount varies** — the official docs mention $250, but some promotions and review sites report $25, $100, or different trial durations (3, 7, or 30 days). Check the current offer at signup time
- **No free tier after trial** — unlike AWS or GCP, UpCloud does not have a permanent free tier. Once the trial ends, you pay for everything
- **Good for indie projects** — simple pricing, no surprise bills from bandwidth, performant servers at competitive prices
- **Terraform and API-first** — UpCloud has a solid developer experience with infrastructure-as-code support
- **Compare with Hetzner, Vultr, and DigitalOcean** — these are the closest competitors in the same market segment

---

**Sources:**
- [UpCloud Free Trial Documentation](https://upcloud.com/docs/getting-started/free-trial/)
- [UpCloud Get Started](https://upcloud.com/get-started-with-upcloud/)
- [UpCloud Data Centers](https://upcloud.com/data-centers/)
- [UpCloud Pricing](https://upcloud.com/pricing/)
- [UpCloud VPS Hosting](https://upcloud.com/vps-hosting/)
- [UpCloud Locations Documentation](https://upcloud.com/docs/getting-started/locations/)
- [GoForTrial: UpCloud Free Trial](https://gofortrial.com/service/upcloud)

## Submission 120: Thunder Compute Free GPU Credits

- Category: Cloud Providers
- Source Link: https://www.thundercompute.com/
- Amount: $20 free credits + up to $50 match on first purchase (up to $70 total value)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Thunder Compute is a Y Combinator-backed GPU cloud platform offering some of the cheapest on-demand GPU pricing available. New users get **$20 in free credits** on signup, plus a **100% match on your first credit purchase up to $50** -- so buy $50, get $100 total. Per-minute billing, no contracts, no minimum spend. Great for ML training, inference, and creative AI workloads.

## Registration (Step-by-Step)

1. Go to [console.thundercompute.com](https://console.thundercompute.com/) and create an account (email + password or OAuth).
2. You receive **$20 in free credits** immediately upon signup.
3. Add a payment method to unlock the credit-matching promotion.
4. Make your first credit purchase (up to $50) and Thunder Compute matches it 100% -- buy $50, get $100 in total credits.
5. Click **Launch GPU** in the console to spin up an instance (~60 seconds).
6. Connect via **SSH**, **Jupyter**, or directly from **VS Code** using their extension.

## Available GPUs & Pricing

| GPU | VRAM | Price |
|-----|------|-------|
| RTX A6000 | 48 GB | $0.27/hr |
| A100 | 80 GB | $0.78/hr |
| H100 PCIe | 80 GB | $1.38/hr |
| Multi-GPU A100 (1-8x) | 80 GB each | $1.79/GPU/hr |
| Multi-GPU H100 (1-8x) | 80 GB each | $2.49/GPU/hr |

**Add-ons:**
- Storage: $0.10/GB/month (100 GB included)
- Snapshots: $0.05/GB/month
- Additional CPU cores: $0.06/core/month

## Free Tier Limits

| Resource | Limit |
|----------|-------|
| Signup credits | $20 (one-time) |
| Credit match | 100% on first purchase, up to $50 match |
| Max free value | $70 total ($20 signup + $50 match) |
| Billing model | Per-minute, pay-as-you-go |
| Contracts | None required |
| Storage included | 100 GB |

## Additional Tips

- **$20 gets you ~25 hours on an A6000** or ~7 hours on a single A100 -- enough for serious prototyping.
- **Pre-built templates** are available for Comfy UI, Ollama, SDXL (Stable Diffusion XL), and Wan2.1 so you can start running workloads immediately.
- **Access methods:** Web console, CLI, and a VS Code extension for seamless remote development.
- **Referral program (beta):** Earn 3% of every dollar your referrals spend as non-transferable credits. No cap on earnings.
- Pricing is ~80% lower than AWS and 2-3x cheaper than Lambda, RunPod, or Paperspace for A100s.
- Data centers are in **North America** only.
- No long-term contracts or upfront commitments.

**Sources:**
- https://www.thundercompute.com/
- https://www.thundercompute.com/pricing
- https://www.thundercompute.com/docs/guides/referral-program
- https://console.thundercompute.com/

## Submission 119: Tencent Cloud Free Tier

- Category: Cloud Providers
- Source Link: https://www.tencentcloud.com/act/pro/FreeTier
- Amount: Multiple always-free services (VPC, Monitor, CloudAudit, 10K TRTC min/mo) + 3-6 month compute/storage/serverless trials + up to $300 welcome voucher
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Tencent Cloud offers a mix of **always-free services** and **time-limited free trials** for new users. The always-free tier includes core networking (VPC, VPN), monitoring (Cloud Monitor), and auditing (CloudAudit). Trial offerings include a CVM compute instance, 50 GB of cloud object storage for 6 months, generous serverless function quotas for 3 months, and 10,000 free TRTC (real-time communication) minutes per month permanently. Tencent Cloud is strongest in Asia-Pacific but operates globally. Good option if you are building for Asian markets or need TRTC/chat/video services.

---

## Registration (Step-by-Step)

1. Go to [tencentcloud.com](https://www.tencentcloud.com) and click **Sign Up**
2. Register with your email or mobile number
3. Complete **identity verification** — choose either individual or enterprise verification
4. For individual: provide ID document; for enterprise: provide business registration
5. Once verified, go to the [Free Tier page](https://www.tencentcloud.com/act/pro/FreeTier) to activate free services
6. Each free trial service must be **individually activated** — they do not auto-enable
7. Some services (like CVM trial) may require adding a payment method

**Important caveats:**
- Free trials are only for **new users of each specific product** — if you have previously used or purchased a service, you cannot get its free trial
- Service suspension due to violations or overdue payment disqualifies you from free tier benefits
- Always-free services remain available as long as your account is active and in good standing

---

## Always-Free Services

| Service | Free Tier Details |
|---|---|
| **VPC (Virtual Private Cloud)** | Free — no charge for VPC, subnets, route tables |
| **Cloud Monitor** | Free — monitoring metrics, alarms, dashboards |
| **CloudAudit** | Free — API call logging and audit trails |
| **ENI (Elastic Network Interface)** | Free — network interfaces for instances |
| **Auto Scaling** | Free — the scaling service itself (you pay for instances launched) |
| **TRTC (Real-Time Communication)** | **10,000 free minutes/month** — audio/video calls, permanently free |
| **Chat (IM)** | Developer Edition free — up to 100 MAU, 1-on-1 chat, group chat, push |

---

## Time-Limited Free Trials

| Service | Free Tier | Duration |
|---|---|---|
| **CVM (Cloud Virtual Machine)** | 1 CPU / 1 GB RAM instance | 3 months |
| **COS (Cloud Object Storage)** | 50 GB STANDARD storage | 6 months (180 days) |
| **SCF (Serverless Cloud Function)** | 1M invocations + 1M GB resource usage + 2 GB outbound traffic/month | 3 months |
| **TencentDB for MySQL** | Free managed database instance | Trial period (varies) |
| **CDN** | Free outbound traffic quota | Trial period |
| **VOD (Video on Demand)** | 10 GB free traffic | 1 year |
| **UGSV (Short Video SDK)** | Unlimited uploads, full features | 28 days |

---

## SCF (Serverless) Free Tier Details

The serverless free tier is generous for the first 3 months:

| Resource | Monthly Free Quota |
|---|---|
| Invocations (event-triggered) | 1,000,000 |
| Invocations (HTTP-triggered) | 1,000,000 |
| Resource usage | 1,000,000 GB-seconds |
| Public network outbound traffic | 2 GB |

Free execution seconds by memory configuration:

| Memory (MB) | Free Seconds/Month |
|---|---|
| 64 | 16,000,000 |
| 128 | 8,000,000 |
| 256 | 4,000,000 |
| 512 | 2,000,000 |
| 1024 | 1,000,000 |

**After 3 months:** transitions to a basic package at $0.06 USD/day. Free quotas do not roll over month-to-month.

---

## COS (Object Storage) Free Tier Caveats

- Only covers **STANDARD storage class** — not STANDARD_IA, ARCHIVE, or DEEP ARCHIVE
- Does **not** cover requests (PUT/GET), data retrieval fees, or traffic fees
- Available in public cloud regions only: Beijing, Shanghai, Guangzhou, Hong Kong, Singapore, Seoul, Tokyo, Silicon Valley, Virginia, Frankfurt, and others
- Financial cloud regions are **excluded**

---

## Vouchers & Bonus Credits

New users may also receive:

| Voucher | Details |
|---|---|
| Welcome voucher | $300 for 30 days (varies by promotion) |
| Media service vouchers | $20-$150 for TRTC, Chat, VOD purchases |

---

## Additional Tips

- **Best for Asia-Pacific** — Tencent Cloud has the strongest presence in China and Southeast Asia with low-latency data centers throughout the region
- **TRTC is the standout free offering** — 10,000 free audio/video minutes per month, permanently, is extremely generous for building communication features
- **Chat SDK free tier** is useful for adding real-time messaging to your app without paying for a third-party service
- **COS free tier is limited** — the 50 GB sounds good but remember requests and traffic are billed separately
- **Serverless (SCF) free tier is very generous** for the first 3 months — 1M invocations and 1M GB-seconds is more than most indie projects need
- **Identity verification is mandatory** — plan for this during registration, it can take 1-2 business days for enterprise verification
- **Interface can be confusing** — Tencent Cloud has multiple portals (intl.cloud.tencent.com vs tencentcloud.com) — use tencentcloud.com for the international version
- **Documentation quality** varies — some docs are machine-translated from Chinese, so cross-reference with API docs when in doubt

---

**Sources:**
- [Tencent Cloud Free Tier](https://www.tencentcloud.com/act/pro/FreeTier)
- [Tencent Cloud COS Free Tier Documentation](https://www.tencentcloud.com/document/product/436/6240)
- [Tencent Cloud SCF Free Tier Documentation](https://www.tencentcloud.com/document/product/583/12282)
- [Tencent Cloud Media & Chat Free Tier](https://www.tencentcloud.com/campaign/media-freetier)
- [TRTC Pricing](https://trtc.io/pricing)
- [Tencent Cloud Pricing Overview](https://www.tencentcloud.com/pricing)
- [Apidog: Free Cloud Service Providers 2026](https://apidog.com/blog/free-cloud-service-providers/)

## Submission 118: SiliconFlow Free Model Inference

- Category: AI API Free Tiers
- Source Link: https://www.siliconflow.com/
- Amount: $1 free credits on signup. Free model access: 50 RPD (upgradeable to 1,000 RPD). Paid models from $0.05/M tokens.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

SiliconFlow is a Chinese AI infrastructure platform offering **free inference for select smaller models** and extremely competitive pricing across 100+ open-source and commercial models. New users get **$1 in free credits** on signup. The platform provides an **OpenAI-compatible API** supporting text, image, video, and audio generation with models from DeepSeek, Qwen, Meta Llama, Mistral, and more. SiliconFlow claims up to **2.3x faster inference** and **32% lower latency** than leading competitors. Free model access is limited to 50 requests/day for unpaid accounts (upgradeable to 1,000/day with a $10 credit purchase).

---

## Registration (Step-by-Step)

1. Go to [siliconflow.com](https://www.siliconflow.com) and click **Sign Up**
2. Create an account (email registration)
3. You automatically receive **$1 in free credits** upon signup
4. Navigate to your dashboard to generate an API key
5. Set the base URL to SiliconFlow's OpenAI-compatible endpoint
6. Start making requests — select free models for zero-cost inference or use credits for premium models

**Important:**
- No credit card required for signup
- Free credits are one-time, not recurring
- Free model rate limits are very restrictive (50 RPD) unless you top up
- The platform is based in China but serves global users

---

## Available AI Models / Services

### Budget Text Models (from $0.05/M tokens)

| Model | Input Price/M Tokens | Output Price/M Tokens | Notes |
|---|---|---|---|
| **Qwen2.5-VL-7B-Instruct** | $0.05 | $0.05 | Vision + video understanding |
| **Meta-Llama-3.1-8B-Instruct** | $0.06 | $0.06 | Multilingual chat |
| **Qwen3-8B** | $0.06 | $0.06 | 100+ languages, 131K context |
| **GLM-4-9B-0414** | $0.086 | $0.086 | Code gen, web design |

### Mid-Range & Premium Models

| Model | Input Price/M Tokens | Output Price/M Tokens |
|---|---|---|
| **DeepSeek-V3.2-Exp** | $0.27 | $0.41 |
| **Qwen2.5-72B** | $0.40 | $0.40 |
| **DeepSeek-R1** | Varies | Varies |

### Flagship Models (2026)

| Model | Category |
|---|---|
| **Qwen3-235B-A22B** | General-purpose flagship |
| **Qwen3-Coder-480B-A35B-Instruct** | Coding specialist |
| **QwQ-32B** | Dedicated reasoning |
| **MiniMax-M2** | Frontier model |

### Image & Multimodal

SiliconFlow also supports image generation models with per-output pricing.

---

## Free Tier Limits

| Feature | Free (L0) | After $10 Purchase |
|---|---|---|
| Free credits | $1 on signup | — |
| Free model daily requests | 50/day | 1,000/day |
| Paid model access | Yes (uses credits) | Yes |
| API compatibility | OpenAI-compatible | OpenAI-compatible |
| Rate limit tier | L0 (most restrictive) | Auto-upgrades based on spend |
| Monthly spending limit | Configurable | Configurable |

### Rate Limit Tiers

| Tier | Monthly Spend | Rate Limits |
|---|---|---|
| L0 | $0 | Most restrictive |
| L1+ | Based on spend | Progressively higher |

Rate limits auto-upgrade when monthly spending reaches the next tier threshold.

---

## Additional Tips

- **OpenAI drop-in replacement** — change `base_url` to SiliconFlow's endpoint and swap your API key. Works with OpenAI SDK, LangChain, LiteLLM, and most compatible tools
- **Buy $10 in credits** to unlock 1,000 daily free model requests (20x the default). This is the single best move for free-tier maximizers
- **Free model requests are separate from credits** — using designated free models does not consume your $1 credit balance
- **Set a monthly spending limit** in your dashboard to prevent unexpected charges
- **Auto-tier upgrades** — your rate limits automatically increase as you spend more per month. No manual tier changes needed
- **Speed advantage** — SiliconFlow benchmarks show 2.3x faster inference than some competitors, which matters for latency-sensitive applications
- **Volume discounts** available upon request for high-volume usage
- **Chinese origin** — the platform documentation is available in both English and Chinese. Some UI elements may default to Chinese
- **Compare with alternatives** — for free-tier inference, also check out Groq (speed-focused), Cerebras (fast), and HuggingFace Inference (breadth)

---

**Sources:**
- [SiliconFlow Platform](https://www.siliconflow.com)
- [SiliconFlow Pricing](https://www.siliconflow.com/pricing)
- [SiliconFlow Models](https://www.siliconflow.com/models)
- [SiliconFlow Rate Limits Documentation](https://docs.siliconflow.com/en/userguide/rate-limits/rate-limit-and-upgradation)
- [SiliconFlow Cheapest LLM Models Guide](https://www.siliconflow.com/articles/en/the-cheapest-LLM-models)
- [SiliconFlow Release Notes](https://docs.siliconflow.cn/en/release-notes/overview)

## Submission 117: Scaleway Startup Program - Up to EUR 36,000

- Category: Cloud Providers
- Source Link: https://www.scaleway.com/en/startup-program
- Amount: EUR 1,000 (Founders, 12mo) to EUR 9,000 (Early Stage, 6mo) to EUR 36,000 (Growth, 12mo) in cloud credits
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Scaleway offers a three-tier startup program with **EUR 1,000 to EUR 36,000 in cloud credits** depending on your stage. Unlike OVHcloud, Scaleway credits **do cover GPU instances** including NVIDIA H100, L40S, and L4 — making this one of the best EU-based programs for AI/ML workloads. Includes technical guidance, a dedicated Customer Success Manager (Early Stage+), beta product access, and a private Slack community with 800+ startup founders. Scaleway is a French cloud provider with data centers in France, Netherlands, and Poland.

---

## Registration (Step-by-Step)

1. **Create a Scaleway account** at [scaleway.com](https://www.scaleway.com) — you must have an account before applying
2. Go to the [Startup Program page](https://www.scaleway.com/en/startup-program/) and click **Apply**
3. Fill out the application: company details, stage, what you are building, infrastructure needs, and current cloud spend
4. If affiliated with an accelerator or VC (Y Combinator, Station F, etc.), **mention it explicitly** — it increases your chances and may unlock higher credit tiers
5. Submit the application to **startup-program@scaleway.com** (or through the web form)
6. Scaleway reviews applications on **monthly rolling cycles** — there is no fixed deadline
7. If selected before the 15th of a month, your program starts on the 1st of the following month
8. Upon acceptance, you receive onboarding with a Program Lead and (for Early Stage+) a Customer Success Manager call

**Important caveats:**
- You must **not be a current paying Scaleway customer** — the program targets new customers
- You only pay if your consumption exceeds your voucher amount in a given month
- Credits are distributed as **monthly vouchers** (not a lump sum), so unused monthly amounts expire
- Maintain a credible digital footprint: up-to-date website, LinkedIn, press coverage
- Be specific about your tech needs and planned infrastructure scale — vague applications get rejected

---

## Program Tiers

| Tier | Credits | Duration | Target | Key Benefits |
|---|---|---|---|---|
| **Founders** | EUR 1,000 total | 12 months | Pre-seed, migrating or initial builds | Launch without cloud costs |
| **Early Stage** | EUR 1,500/month (EUR 9,000 total) | 6 months | Going to market | Community support, co-marketing, dedicated CSM |
| **Growth** | EUR 3,000/month (EUR 36,000 total) | 12 months | Scaling up (seed to Series A) | Advanced technical support, full benefits |

---

## Available AI/ML & GPU Services

Scaleway has heavily invested in AI infrastructure. Credits can be used on:

| Service | Description |
|---|---|
| **NVIDIA H100 GPU Instances** | Access to Nabuchodonosor supercomputer (1,016 H100 GPUs) |
| **NVIDIA L40S GPU Instances** | Inference and mixed workloads |
| **NVIDIA L4 GPU Instances** | Cost-efficient inference |
| **Managed Inference** | Deploy pre-trained models as API endpoints |
| **GPU Cluster On Demand** | Rent multi-GPU clusters for large training jobs |
| **Instances** | General-purpose, CPU-optimized, memory-optimized VMs |
| **Managed Kubernetes (Kapsule)** | Container orchestration |
| **Managed Databases** | PostgreSQL, MySQL, Redis |
| **Object Storage** | S3-compatible storage |
| **Serverless Functions** | Event-driven compute |
| **Serverless Containers** | Deploy containers without managing servers |

---

## NVIDIA Inception Integration

Scaleway startup program members are eligible to join **NVIDIA Inception** — NVIDIA's own startup accelerator. Benefits include:

| Benefit | Details |
|---|---|
| NVIDIA AI Enterprise software | NeMo framework, pretrained LLMs |
| NVIDIA RAPIDS | Accelerated data science |
| NVIDIA Triton Inference Server | High-performance model serving |
| NVIDIA TensorRT-LLM | Optimized LLM inference |
| Marketplace access | Pre-built AI containers and models |

---

## Partner Perks

Scaleway startup program includes over **EUR 250,000 in total partner perks** from various companies (specific partners vary by cohort). Additional benefits:

| Benefit | Details |
|---|---|
| Private Slack community | 800+ startup founders |
| Masterclasses & webinars | Technical and business content |
| Early beta access | Test new Scaleway products before GA |
| Co-marketing | Case studies, event features, blog posts |
| Event opportunities | Speaking slots, booth access |

---

## Additional Tips

- **Best EU program for AI/ML** — unlike OVHcloud, Scaleway credits cover GPU instances (H100, L40S, L4)
- **Show growth potential** in your application — Scaleway wants to support startups that will outgrow the free tier and become paying customers
- **Monthly voucher model** means you need to plan usage carefully; unused monthly credits do not roll over
- **Data centers** are in Paris, Amsterdam, and Warsaw — good for EU-based workloads with low latency
- **No equity, no exclusivity** — Scaleway does not take a stake in your company
- **Combine with NVIDIA Inception** for additional GPU credits and AI tooling
- **French tech ecosystem** — strong integration with Station F and French startup networks

---

**Sources:**
- [Scaleway Startup Program](https://www.scaleway.com/en/startup-program/)
- [Scaleway Early Stage Program](https://www.scaleway.com/en/startup-program/early-stage-program/)
- [Scaleway Growth Stage Program](https://www.scaleway.com/en/startup-program/growth-stage-program/)
- [Scaleway GPU Instances](https://www.scaleway.com/en/gpu-instances/)
- [Scaleway H100 Instances](https://www.scaleway.com/en/h100/)
- [Scaleway AI Solutions](https://www.scaleway.com/en/ai-solutions/)
- [Scaleway GPU Pricing](https://www.scaleway.com/en/pricing/gpu/)
- [NVIDIA Blog: Scaleway + NVIDIA Inception](https://blogs.nvidia.com/blog/scaleway-european-startups-inception/)
- [xRaise: Scaleway for Startups](https://xraise.ai/blog/scaleway-for-startups-is-it-the-right-choice/)
- [xRaise: Scaleway Promo Code for Startups](https://xraise.ai/blog/scaleway-promo-code-for-startups-save-up-to-e36000/)

## Submission 116: Saturn Cloud Free GPU Plan - 150 Hours/Month

- Category: Free GPU & Compute
- Source Link: https://saturncloud.io/plans/saturn_cloud_plans
- Amount: 150 hours/month compute (GPU T4 16GB + CPU, shared pool)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Saturn Cloud offers a **free Hosted plan with 150 hours of compute per month**, covering both CPU and GPU instances. You get access to **NVIDIA T4 GPUs (16 GB VRAM)** through browser-based JupyterLab, along with Dask clusters for distributed computing. No credit card required. The platform supports Python and R, integrates with popular IDEs like VS Code and PyCharm via SSH, and includes persistent storage so your work survives between sessions.

---

## Registration (Step-by-Step)

1. Go to [saturncloud.io](https://saturncloud.io) and click **Start for free**
2. You'll be redirected to the hosted registration page at `app.community.saturnenterprise.io`
3. Sign up with your email address or a supported SSO method (Google, GitHub)
4. Confirm your email address
5. Once logged in, create a new Jupyter Server resource and select a **T4-XLarge GPU** instance type under the free plan
6. Click **Start** to launch your GPU-powered notebook environment

**Important:**
- No credit card is required for the free plan
- The free plan was previously 10 hours GPU + 3 hours Dask per month, then upgraded to 30 hours, and is now at **150 hours/month**
- Hours are shared across GPU and CPU instances -- using a GPU instance consumes from the same 150-hour pool
- Hours reset at the beginning of each calendar month

---

## GPU Hardware & Specs

| Resource | Specification |
|---|---|
| **GPU** | NVIDIA Tesla T4 |
| **GPU VRAM** | 16 GB GDDR6 |
| **GPU Architecture** | Turing (Tensor Cores for FP16/INT8) |
| **CPU Cores (GPU instance)** | 4 vCPUs |
| **RAM (GPU instance)** | 16 GB |
| **CPU-only instances** | Up to 64 GB RAM |
| **Dask Workers (free tier)** | 3 workers, 4 cores + 16 GB RAM + 1 GPU each |

---

## Free Tier Limits

| Resource | Limit |
|---|---|
| **Monthly compute hours** | 150 hours (GPU + CPU combined) |
| **GPU instance type** | T4-XLarge only |
| **CPU instance type** | Large CPU only |
| **Dask clusters** | Included (consumes from same hour pool) |
| **Persistent storage** | Included |
| **Auto-shutdown** | Workspaces auto-shutdown when idle |
| **Concurrent resources** | Limited (exact number not published) |

---

## Features Included in Free Plan

- **JupyterLab** notebooks with GPU support
- **RStudio** support for R users
- **VS Code and PyCharm** connection via SSH
- **Dask** distributed computing clusters
- **Git integration** for version-controlled workflows
- **Shared folders** for collaboration
- **Pre-installed libraries**: PyTorch, TensorFlow, RAPIDS, LightGBM, XGBoost, and standard data science stack
- **Docker image customization** for custom environments
- **Jobs and Deployments** for running scheduled tasks and serving models
- **Background execution** so training continues after you close the browser

---

## Additional Tips

- **Idle auto-shutdown saves hours**: Saturn Cloud automatically shuts down idle workspaces, so you won't waste your 150-hour quota by forgetting to stop a session
- **Use Dask for distributed workloads**: The free tier includes Dask clusters, which is rare for free GPU plans. Great for data processing at scale
- **RAPIDS acceleration**: Saturn Cloud has deep integration with NVIDIA RAPIDS for GPU-accelerated Pandas, scikit-learn, and graph analytics
- **Chain with other free tiers**: Export models trained on Saturn Cloud and deploy elsewhere (e.g., Hugging Face Spaces, Modal, or Beam Cloud)
- **Paid upgrade path**: The Pro plan adds pay-as-you-go access to larger GPU instances including H100 and H200 at $2.95/hour
- **Enterprise options**: Hosted Pro and VPC deployments on AWS available for teams needing more control
- **Frequent disconnections**: Some users report occasional disconnections on the free tier; save checkpoints frequently

---

**Sources:**
- [Saturn Cloud Plans and Pricing](https://saturncloud.io/plans/saturn_cloud_plans/)
- [Saturn Cloud 150 Free Hours Announcement](https://saturncloud.io/blog/150-free-hours/)
- [Saturn Cloud 30 Free Hours Blog Post](https://saturncloud.io/blog/saturn-cloud-offers-free-compute-time/)
- [Saturn Cloud Free GPU Landing Page](https://saturncloud.io/try/gpus-on-saturncloud/)
- [Free Cloud GPU Comparison 2026 - AI Multiple](https://research.aimultiple.com/free-cloud-gpu/)
- [Free GPU Cloud Trials 2026 - GMI Cloud](https://www.gmicloud.ai/blog/where-can-i-get-free-gpu-cloud-trials-in-2026-a-complete-guide)

## Submission 115: Render Free Hosting Tier

- Category: Cloud Providers
- Source Link: https://render.com/docs/free
- Amount: 750 free instance hours/month, free static hosting, 1 GB PostgreSQL (30-day expiry), free Key Value (ephemeral)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Render offers a solid free tier for deploying web apps, static sites, PostgreSQL databases, and Redis (Key Value) instances. No credit card required. Auto-deploys from Git (GitHub/GitLab). Great for side projects, portfolio sites, and MVPs -- but be aware of idle spin-down behavior and the 30-day PostgreSQL expiration.

## Registration (Step-by-Step)

1. Go to [render.com](https://render.com/) and sign up with GitHub, GitLab, Google, or email.
2. No credit card is required for the free tier.
3. Create a new **Web Service**, **Static Site**, **PostgreSQL**, or **Key Value** instance from the dashboard.
4. Connect your Git repository for automatic deployments on push.
5. Render auto-detects your framework and builds/deploys your app.

## Free Tier Limits

| Service | Free Tier Details |
|---------|------------------|
| Web Services | 750 free instance hours/month per workspace; spins down after 15 min of inactivity (~1 min cold start) |
| Static Sites | Unlimited static sites; includes CDN, custom domains, TLS |
| PostgreSQL | 1 GB storage; **expires 30 days after creation** (14-day grace period to upgrade); 1 per workspace; no backups |
| Key Value (Redis) | 1 per workspace; **ephemeral only** -- data lost on restart; in-memory, no disk persistence |
| Bandwidth | Included with workspace plan; overage suspends services if no payment method on file |
| Persistent Disk | Not available on free tier |
| SSH Access | Not available on free tier |
| Scaling | Not available on free tier |

## Additional Tips

- **The biggest gotcha: Free PostgreSQL databases expire after 30 days.** You get a 14-day grace period to upgrade or export your data, then it is deleted. Plan accordingly -- this is not a permanent free database.
- **Spin-down behavior:** Free web services go idle after 15 minutes without traffic. First request after idle takes ~60 seconds. Not suitable for APIs that need instant responses.
- **Ephemeral filesystem:** Any files written to disk during runtime are lost on every redeploy. Use external storage (S3, Cloudflare R2) for persistent files.
- **No SMTP:** Ports 25, 465, 587 are blocked on free tier. Use a third-party email service (SendGrid, Mailgun, Resend).
- **Pull Request Previews** are supported -- each PR gets its own preview URL.
- **Custom domains** with automatic TLS are free.
- **Cron jobs** are available but count against your instance hours.
- For a permanent free database, consider pairing Render's free web service with a free Supabase, Neon, or PlanetScale database instead.

**Sources:**
- https://render.com/docs/free
- https://render.com/blog/free-tier
- https://render.com/docs/postgresql-refresh

## Submission 114: Railway Free Trial Credits

- Category: Cloud Providers
- Source Link: https://railway.com/pricing
- Amount: $5 one-time trial credit (30 days) + $1/month ongoing on Free plan
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Railway is a modern PaaS for deploying apps, databases, and services from Git with minimal config. New accounts get a **$5 one-time trial credit** for 30 days, then revert to the Free plan with **$1/month** in ongoing credit. Supports Node.js, Python, Go, Rust, Ruby, and more. Also offers one-click Postgres, MySQL, Redis, and MongoDB deployments.

## Registration (Step-by-Step)

1. Go to [railway.com](https://railway.com/) and sign up with **GitHub** (recommended for full network access) or email.
2. You immediately receive **$5 in free trial credits**, valid for 30 days.
3. Deploy a project from a Git repo, template, or Docker image.
4. After the trial ends (30 days or $5 spent), your account moves to the **Free plan** with $1/month credit.
5. To get full features, upgrade to the **Hobby plan** ($5/month with $5 usage credit included).

**Important:** Signing up with a verified GitHub account grants **full network access** during the trial. Unverified accounts get restricted outbound networking and limited ports.

## Free Tier Limits

| Resource | Trial (30 days) | Free Plan (ongoing) |
|----------|-----------------|---------------------|
| Credits | $5 one-time | $1/month (no rollover) |
| RAM per service | 1 GB | 0.5 GB |
| vCPU | Shared | 1 vCPU |
| Ephemeral Storage | -- | 1 GB |
| Volume Storage | -- | 0.5 GB |
| Services per project | 5 | -- |
| Image Size | -- | 4 GB |
| Replicas | -- | 0 |
| Network | Full (GitHub verified) / Limited (unverified) | -- |

## Pricing (usage-based)

| Resource | Cost |
|----------|------|
| RAM | $10/GB/month |
| vCPU | $20/vCPU/month |
| Network Egress | $0.05/GB |
| Volume Storage | $0.15/GB/month |

## Additional Tips

- **$5 goes further than you think** for small apps. A typical Node.js API with a Postgres database uses ~$2-3/month on Railway.
- **Credit does not roll over** on the Free plan -- use it or lose it each month.
- **GitHub verification matters:** Sign up with a GitHub account that has some history (repos, activity). Unverified accounts get restricted networking, which can break apps that make external API calls.
- **Stateful volumes are deleted 30 days after trial credits expire.** Export your data before the trial ends if you are not upgrading.
- Railway supports **one-click database deployments** (Postgres, MySQL, Redis, MongoDB) that share your credit allocation.
- The **Hobby plan at $5/month** is a natural upgrade path -- it includes $5 in usage credit, so you only pay for overages.
- Railway has excellent **monorepo support** and can auto-detect services in a multi-service repo.
- **Image retention** on the Free plan is only 24 hours after deployment removal.

**Sources:**
- https://railway.com/pricing
- https://docs.railway.com/reference/pricing/free-trial
- https://docs.railway.com/reference/pricing/plans
- https://blog.railway.com/p/free-plan

## Submission 113: Pollinations AI Free API (No Signup Required)

- Category: AI API Free Tiers
- Source Link: https://pollinations.ai/
- Amount: Free Flux images (unlimited). Registered free tier: 1.5 Pollen/week. Anonymous: ~1 req/15 sec.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Pollinations is an open-source AI platform based in Berlin that lets you generate images, text, audio, and video via API -- with zero signup required for basic use. You can hit their endpoints directly with a URL. Free Flux image generation is unlimited and always free. For text and other models, you get Pollen credits (free tier: 1.5 Pollen/week). The API supports Claude, GPT-5, Gemini, DeepSeek, Llama, Mistral, and more.

## Registration (Step-by-Step)

**No-signup usage (anonymous):**
1. Generate an image: open `https://image.pollinations.ai/prompt/a%20cat%20coding` in your browser. Done.
2. Generate text: `GET https://text.pollinations.ai/your-prompt-here`
3. No API key, no account, no headers needed.

**Registered usage (more features + higher limits):**
1. Go to [enter.pollinations.ai](https://enter.pollinations.ai) and sign up.
2. Get a **Publishable key (pk_)** for client-side use or a **Secret key (sk_)** for server-side use.
3. Use the unified endpoint at `gen.pollinations.ai` with your key.
4. You start at the **Spore tier** with 1.5 Pollen/week for free.

**Caveat:** Publishable keys have limited protection (no Turnstile yet). Never expose secret keys in client-side code.

## Available AI Models / Services

### Image Generation
- Flux (default, always free)
- GPT Image
- Seedream
- Kontext
- Nanobanana

### Text Generation
- Claude (default)
- GPT-5
- Gemini
- DeepSeek
- Llama
- Mistral
- Qwen
- Specialized agents: **nomnom** (web research), **polly** (coding)

### Audio
- Text-to-Speech with 30+ voices (OpenAI-compatible)
- Speech-to-Text

### Video (Alpha)
- Seedance
- Veo

## Free Tier Limits

| Tier | Cost | Rate Limit | Notes |
|---|---|---|---|
| Anonymous (no key) | Free | ~1 req/15 sec | Image generation only |
| Publishable key (pk_) | Free | 1 Pollen/hr per IP+key | Client-side demos |
| Spore (registered, free) | Free | 1.5 Pollen/week | Default registered tier |
| Seed (registered) | Paid | 3 Pollen/hr | For active devs |
| Flower (registered) | Paid | 10 Pollen/hr | Requires app submission |
| Secret key (sk_) | Paid | No rate limits | Server-side only |

- **Flux images are always free** regardless of tier.
- 1 Pollen ~ $1 in compute costs.

## API Endpoints

| Endpoint | Description |
|---|---|
| `gen.pollinations.ai/image/{prompt}` | Image generation |
| `gen.pollinations.ai/text/{prompt}` | Text generation |
| `gen.pollinations.ai/audio/{text}` | Text-to-speech |
| `gen.pollinations.ai/v1/audio/speech` | OpenAI TTS compatible |
| `image.pollinations.ai/prompt/{prompt}` | Legacy image endpoint |
| `text.pollinations.ai/{prompt}` | Legacy text endpoint |

## Additional Tips

- **Simplest free AI API that exists:** Just construct a URL and make a GET request. No SDK, no auth, no setup. Perfect for prototypes, demos, and hobby projects.
- **OpenAI-compatible TTS:** The `/v1/audio/speech` endpoint works with existing OpenAI audio code.
- **Bring Your Own Pollen:** Developers can let their users pay for compute, so the developer pays zero.
- **Cloudflare caching:** Repeated image generations for the same prompt are cached and served instantly.
- **Open source:** The entire platform is open-source at [github.com/pollinations/pollinations](https://github.com/pollinations/pollinations).
- **500+ community projects** already built on top of this API.
- **MCP server support:** Integrates with Claude and other AI assistants as a tool.
- **Not for production at scale:** The free tier is designed for prototyping and community projects. For production, consider the paid tiers or self-hosting.

**Sources:**
- https://pollinations.ai
- https://github.com/pollinations/pollinations
- https://enter.pollinations.ai/api/docs
- https://github.com/pollinations/pollinations/blob/master/enter.pollinations.ai/POLLEN_FAQ.md

## Submission 112: OVHcloud Startup Program - Up to EUR 100,000

- Category: Cloud Providers
- Source Link: https://startup.ovhcloud.com/en
- Amount: EUR 10,000 (START) to EUR 100,000 (SCALE) in cloud credits + 6-20h engineering consultation, 12-month program
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

OVHcloud runs a tiered startup program offering **EUR 10,000 to EUR 100,000 in free public cloud credits** for 12 months, plus up to 20 hours of 1-on-1 engineering consultation. No equity taken, no exclusivity required. Credits cover OVHcloud Public Cloud and Hosted Private Cloud products. OVHcloud is a European-sovereign cloud provider with GDPR-compliant infrastructure, making it a strong pick if you need EU data residency. The program also includes access to a community platform (Whaller), event visibility, and alumni networking.

---

## Registration (Step-by-Step)

1. Go to [startup.ovhcloud.com](https://startup.ovhcloud.com/en/) and verify you are on the correct country website matching your business address
2. Click **"Apply"** on the homepage
3. Log in with your existing OVHcloud account or create a new one (you cannot switch accounts later)
4. Fill out the application form: company name, website, description of your project, funding stage, team size, and infrastructure needs
5. Submit the application
6. The selection committee reviews applications daily and responds within **7 days**
7. If accepted, you sign the program contract — vouchers activate within **48 hours** of signing

**Important caveats:**
- You must have at least a **POC or MVP** ready to deploy on OVHcloud infrastructure
- Company must be **less than 5 years old**, under **50 employees**, and under **EUR 10M** in revenue
- Non-professional entities (NGOs, associations) are **not eligible**
- Previous program participants are **not eligible** for the same tier again
- Cryptocurrency mining, Black Hat SEO, traffic exchange bots (Hitleap, Jingling) are **prohibited**
- Rolling applications throughout the year with limited quarterly slots
- Credits are use-it-or-lose-it within the 12-month program period

---

## Program Tiers

| Tier | Credits | Engineering Hours | Target |
|---|---|---|---|
| **START** (Startups) | EUR 10,000 | 6 hours of 1-on-1 support | Pre-seed & seed stage, monthly cloud spend in hundreds |
| **SCALE** (Scaleups) | Up to EUR 100,000 | Up to 20 hours with solution architects | Post-Series A, monthly cloud spend in thousands |
| **Enablers** | Fast-track access for portfolio | Co-hosting & content collaboration | VCs, accelerators, ecosystem partners |

---

## Services Covered by Credits

| Service | Covered? |
|---|---|
| Public Cloud (compute, storage, networking) | Yes |
| Hosted Private Cloud | Yes |
| Managed Kubernetes | Yes |
| Managed Databases | Yes |
| AI Training (OVHcloud AI Training) | Yes |
| AI Deploy (model deployment) | Yes |
| **A100, H100, L4, L40s GPU instances** | **No — excluded** |
| Bare Metal Cloud | **No** |
| Web Cloud products | **No** |

**Key limitation:** GPU instances (A100, H100, L4, L40s) are explicitly excluded from startup program vouchers. If you need GPU access for AI/ML training, you will need to pay separately or look at other providers.

---

## Available AI/ML Services (pay-as-you-go, credits apply to non-GPU)

| Service | Description |
|---|---|
| **AI Training** | Train models on managed infrastructure (CPU-based covered by credits) |
| **AI Deploy** | Deploy and serve ML models as APIs |
| **AI Notebooks** | Managed JupyterHub for data science |
| **NVIDIA NGC** | Pre-built AI frameworks on OVHcloud GPU Cloud |
| **L40S Cloud GPU** | Generative AI inference, DL training, 3D rendering (not covered by credits) |

---

## Additional Tips

- **EU data sovereignty** is OVHcloud's strongest selling point — if your customers or regulations require data to stay in Europe, this program is ideal
- **No bandwidth overage charges** — OVHcloud includes generous bandwidth in pricing, unlike AWS/GCP/Azure
- **Open-source friendly** — no vendor lock-in, compatible with standard APIs (OpenStack, S3-compatible storage)
- **Regional priority:** Germany, Austria, and Switzerland-based startups get priority consideration
- **Partner referrals** from accelerators or VCs may help your application, though the FAQ says event codes carry "no specific advantages"
- **Alumni community** — after the 12-month program ends, you join an alumni network with continued event access
- **Compare with Scaleway** — another EU-sovereign cloud with a startup program (up to EUR 36,000), which does include GPU access

---

**Sources:**
- [OVHcloud Startup Program](https://startup.ovhcloud.com/en/)
- [OVHcloud Startup Program for Startups](https://startup.ovhcloud.com/en/startups/)
- [OVHcloud Startup Program for Scaleups](https://startup.ovhcloud.com/en/scaleups/)
- [OVHcloud Startup Program FAQ](https://startup.ovhcloud.com/en/faq-support/)
- [OVHcloud Available Products for Credits](https://help.ovhcloud.com/csm/en-startup-program-available-products?id=kb_article_view&sysparm_article=KB0073088)
- [OVHcloud GPU Cloud](https://www.ovhcloud.com/en/public-cloud/gpu/)
- [OVHcloud AI Training](https://us.ovhcloud.com/public-cloud/ai-training/)
- [StartUs Magazine: OVHcloud Startup Program 2025](https://magazine.startus.cc/ovhcloud-startup-program-2025/)
- [CloudCredits.io: OVHcloud](https://cloudcredits.io/providers/ovh/programs/ovhcloud-startup-program)

## Submission 111: OVHcloud AI Endpoints Free Tier

- Category: AI API Free Tiers
- Source Link: https://www.ovhcloud.com/en/public-cloud/ai-endpoints
- Amount: Free trial credit (1 month). 4 permanently free models (SDXL, Riva TTS, Qwen3Guard). 40+ models total.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

OVHcloud, a European cloud provider, offers serverless access to 40+ open-source AI models via API. Several models are completely free (image generation, TTS, content moderation), while LLMs are available at low per-token pricing with a free trial credit for new accounts. All data is hosted in the EU with GDPR compliance -- no data stored or used for model training. No API key required for playground access.

## Registration (Step-by-Step)

**Anonymous playground (no signup):**
1. Go to [ovhcloud.com/en/public-cloud/ai-endpoints/catalog/](https://www.ovhcloud.com/en/public-cloud/ai-endpoints/catalog/).
2. Click **Try** on any model to test it in the playground.
3. Rate-limited per IP, per minute, per model. No API key needed.

**Authenticated API access (free trial):**
1. Go to [ovhcloud.com](https://www.ovhcloud.com) and create a free Public Cloud account.
2. You'll receive a **free credit valid for one month** that covers AI Endpoints usage.
3. Navigate to the **OVHcloud Manager** and create a Public Cloud project.
4. Go to **AI Endpoints** and generate an API key.
5. Use the API with your key for higher rate limits.

**Caveat:** The free trial credit is valid for only 1 month. After that, you pay per token. Some models (SDXL, Riva TTS, LLM Guard) remain free permanently.

## Available AI Models / Services

### Large Language Models
| Model | Input Price | Output Price |
|---|---|---|
| Llama 3.1 8B Instruct | EUR 0.10/M tokens | EUR 0.10/M tokens |
| Meta Llama 3.3 70B Instruct | EUR 0.67/M tokens | EUR 0.67/M tokens |
| Mistral 7B Instruct v0.3 | EUR 0.10/M tokens | EUR 0.10/M tokens |
| Mistral Nemo Instruct | EUR 0.13/M tokens | EUR 0.13/M tokens |

### Reasoning LLMs
| Model | Input Price | Output Price |
|---|---|---|
| GPT-OSS 20B | EUR 0.04/M tokens | EUR 0.15/M tokens |
| GPT-OSS 120B | EUR 0.08/M tokens | EUR 0.40/M tokens |
| Qwen3 32B | EUR 0.08/M tokens | EUR 0.23/M tokens |

### Code LLM
| Model | Input Price | Output Price |
|---|---|---|
| Qwen3 Coder 30B A3B | EUR 0.06/M tokens | EUR 0.22/M tokens |

### Vision LLMs
| Model | Input Price | Output Price |
|---|---|---|
| Mistral Small 3.2 24B | EUR 0.09/M tokens | EUR 0.28/M tokens |
| Qwen 2.5 VL 72B | EUR 0.91/M tokens | EUR 0.91/M tokens |

### Free Models (no charge)
| Model | Category |
|---|---|
| Stable Diffusion XL Base v1.0 | Image Generation |
| NVIDIA Riva TTS (EN, IT, DE, ES) | Text-to-Speech |
| Qwen3Guard Gen 8B | Content Moderation |
| Qwen3Guard Gen 0.6B | Content Moderation |

### Audio Analysis
| Model | Price |
|---|---|
| Whisper Large v3 Turbo | EUR 0.00001278/sec |
| Whisper Large v3 | EUR 0.00004083/sec |

### Embeddings
| Model | Price |
|---|---|
| BGE Multilingual Gemma2 | EUR 0.01/M tokens |
| BGE M3 | EUR 0.01/M tokens |

## Free Tier Limits

| Access Type | Rate Limit | Cost |
|---|---|---|
| Anonymous (playground) | Per IP, per minute, per model | Free |
| Authenticated (API key) | Per project, per minute, per model | Free trial credit (1 month) |
| SDXL Image Generation | Rate-limited | Always free |
| NVIDIA Riva TTS | Rate-limited | Always free |
| Qwen3Guard Moderation | Rate-limited | Always free |

## Additional Tips

- **EU data sovereignty:** All processing happens in EU data centers. OVHcloud guarantees your data is never stored or used for training. If GDPR compliance matters to you, this is a strong option.
- **Permanently free models:** SDXL for image generation, NVIDIA Riva for TTS (4 languages), and Qwen3Guard for content moderation are free with no trial expiration.
- **OpenAI-compatible API:** Many OVHcloud AI Endpoints use OpenAI-compatible formats, making integration straightforward with existing SDKs.
- **Three API tiers:** Base API (no commitment, standard speed), Fast API (committed, ultra-fast), and Batch API (cheapest, async processing).
- **LiteLLM and LlamaIndex support:** Official integrations with LiteLLM and LlamaIndex for easy switching between providers.
- **Competitive pricing:** Llama 3.1 8B at EUR 0.10/M tokens and GPT-OSS 20B at EUR 0.04/M input tokens are very competitive.
- **SSL/TLS encryption:** All API calls are encrypted. API key authentication for production use.

**Sources:**
- https://www.ovhcloud.com/en/public-cloud/ai-endpoints/
- https://www.ovhcloud.com/en/public-cloud/ai-endpoints/catalog/
- https://www.ovhcloud.com/en/public-cloud/free-ai-api/
- https://corporate.ovhcloud.com/en/newsroom/news/ai-endpoints/
- https://docs.litellm.ai/docs/providers/ovhcloud

## Submission 110: OpenRouter Free Models

- Category: AI API Free Tiers
- Source Link: https://openrouter.ai/collections/free-models
- Amount: 29+ free models. 20 req/min, 50 req/day (or 1,000/day with $10+ credits on account).
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

OpenRouter gives you a single OpenAI-compatible API endpoint to access 29+ free AI models -- including DeepSeek, Gemini, Llama 3.3, Qwen3, Mistral, and more. No credit card needed. Free models are community-funded and rotated; append `:free` to any supported model ID to use it at zero cost.

## Registration (Step-by-Step)

1. Go to [openrouter.ai](https://openrouter.ai) and click **Sign Up**.
2. Create an account with email or sign in via Google/GitHub.
3. Navigate to **Keys** in your dashboard and generate an API key.
4. Use the API key with the base URL `https://openrouter.ai/api/v1` -- it's OpenAI-compatible, so most SDKs work out of the box.
5. To use free models, append `:free` to the model ID (e.g., `meta-llama/llama-3.3-70b-instruct:free`).
6. Alternatively, use the **Free Models Router** (`openrouter/free`) to automatically pick a random free model that supports your request's features (vision, tool calling, etc.).

**Caveat:** Free models have lower reliability and may go offline without notice. Don't depend on a specific free model for production.

## Available AI Models / Services

As of February 2026, 29 free models are available, including:

| Model | Provider | Context Length |
|---|---|---|
| Qwen3 Next 80B A3B Instruct | Alibaba | 262K |
| Qwen3 Coder 480B A35B | Alibaba | 262K |
| Step 3.5 Flash | StepFun | 256K |
| Nemotron 3 Nano 30B A3B | NVIDIA | 256K |
| Llama 3.3 70B Instruct | Meta | 128K |
| Gemma 3 27B | Google | 131K |
| Hermes 3 405B | Nous Research | 131K |
| Mistral Small 3.1 24B | Mistral | 128K |
| Solar Pro 3 | Upstage | 128K |
| GPT-OSS 120B / 20B | OpenAI | 131K |
| GLM 4.5 Air | Z.ai | 131K |
| Qwen3 VL 30B / 235B (Vision) | Alibaba | 131K |
| Trinity Mini / Large Preview | Arcee AI | 131K |
| Nemotron Nano 12B/9B v2 | NVIDIA | 128K |
| Gemma 3 4B / 12B | Google | 32K-131K |
| LFM-2.5 variants | Liquid | varies |
| Dolphin Mistral | Cognitive Computations | varies |

The roster changes frequently. Check [openrouter.ai/collections/free-models](https://openrouter.ai/collections/free-models) for the live list.

## Free Tier Limits

| Condition | Rate Limit | Daily Limit |
|---|---|---|
| No purchased credits | 20 req/min | 50 req/day |
| $10+ in credits purchased (ever) | 20 req/min | 1,000 req/day |

- Limits are enforced **globally across all your API keys** -- creating multiple keys won't help.
- Hitting limits returns a `429` error. Daily limits reset at midnight UTC.
- Cloudflare DDoS protection will block requests that dramatically exceed reasonable usage.

## Additional Tips

- **Boost your daily limit 20x:** Add just $10 in credits (you can use them for paid models later) and your free model daily cap jumps from 50 to 1,000 requests.
- **Free Models Router:** Use model ID `openrouter/free` to auto-select a working free model. It intelligently filters for models supporting your request's features (vision, tools, structured output).
- **Monitor usage:** `GET https://openrouter.ai/api/v1/key` returns your remaining credits, daily/weekly/monthly usage, and rate limit status.
- **Negative balance trap:** If your account somehow goes negative, even free models will return `402 Payment Required`.
- **OpenAI SDK compatible:** Just set `base_url="https://openrouter.ai/api/v1"` and your existing OpenAI code works.
- **Not for production:** Free models have variable uptime and may be removed. Use them for prototyping and testing.

**Sources:**
- https://openrouter.ai/collections/free-models
- https://openrouter.ai/docs/api/reference/limits
- https://openrouter.ai/docs/guides/routing/model-variants/free
- https://openrouter.ai/docs/guides/routing/routers/free-models-router
- https://costgoat.com/pricing/openrouter-free-models

## Submission 109: OpenAI NextGenAI Consortium ($50M for Universities)

- Category: Education & Student
- Source Link: https://openai.com/index/introducing-nextgenai
- Amount: Varies by institution; total $50M pool. Example: Texas A&M received $50K cash + $100K API credits.
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

OpenAI's **NextGenAI** is a first-of-its-kind consortium with 15 leading research institutions, backed by **$50 million in research grants, compute funding, and API credits**. The program gives partner universities prepaid access to OpenAI's platform, enabling faculty and students to train, fine-tune, and experiment with AI models across disciplines. Individual students and researchers access credits **through their institution** — there is no direct individual application to NextGenAI. Launched March 2025.

---

## Registration (Step-by-Step)

NextGenAI works through **institutional partnerships**, not direct student applications. Here is how to access the resources:

1. **Check if your university is a founding partner** (see list below)
2. Contact your university's **AI research office, CS department, or provost's office** to ask about NextGenAI credit allocation
3. Faculty typically apply through internal channels — each university has its own distribution process for OpenAI API credits
4. Students usually gain access through:
   - **Faculty-sponsored research projects** that receive NextGenAI funding
   - **University-wide AI access programs** (e.g., MIT gives students and faculty direct API access)
   - **Course-integrated access** (e.g., Howard University embeds AI into curricula)
5. If your university is NOT a partner, check the **OpenAI Researcher Access Program** — a separate program where individual academics can apply for API credits at [openai.com/form/researcher-access-program](https://openai.com/form/researcher-access-program/)

**Important:** There is no public application form for NextGenAI itself. Credits are distributed at the institutional level.

---

## What You Get

| Resource | Details |
|---|---|
| **API Credits** | Prepaid access to OpenAI's models (GPT-4, GPT-4o, etc.) for research and education |
| **Compute Funding** | Resources for training, fine-tuning, and running AI models |
| **Research Grants** | Cash grants for research projects (amounts vary by institution) |
| **Technical Support** | Access to OpenAI's technical resources and expertise |

### Known Funding Examples

- **Texas A&M:** $50,000 cash grant + $100,000 in API credits
- **MIT:** Student and faculty API access + compute resources for model training and fine-tuning
- **Howard University:** Curriculum integration funding + operational AI tools
- **Harvard + Boston Children's Hospital:** Dedicated funding for rare disease diagnosis research
- **Ohio State:** Resources for digital health, therapeutics, and manufacturing AI research

Total $50M is distributed across all 15 founding institutions, but per-institution allocations are not publicly disclosed beyond these examples.

---

## Founding Partner Institutions

| Institution | Focus Area |
|---|---|
| **California Institute of Technology (Caltech)** | Research |
| **California State University System** | Teaching & access (largest US public university system) |
| **Duke University** | Research |
| **University of Georgia** | Research |
| **Harvard University** | Medical AI research (with Boston Children's Hospital) |
| **Howard University** | Curriculum integration, student experience |
| **Massachusetts Institute of Technology (MIT)** | Student/faculty API access, model training |
| **University of Michigan** | Research partnership |
| **University of Mississippi** | Research |
| **The Ohio State University** | Digital health, therapeutics |
| **University of Oxford** | Research (only non-US founding member) |
| **Sciences Po** | Social sciences AI research |
| **Texas A&M University** | Generative AI literacy, research |
| **Boston Children's Hospital** | Rare disease diagnosis |
| **Boston Public Library** | Public access and education |

---

## Eligibility & Verification

- **Faculty and researchers** at partner institutions can apply through internal university channels
- **Students** access credits through faculty-sponsored projects or university-wide programs
- **No direct individual application** to NextGenAI exists
- **Non-partner institutions:** Use the separate [OpenAI Researcher Access Program](https://openai.com/form/researcher-access-program/) for individual API credit applications
- The consortium is **US-focused initially** but includes Oxford (UK) and Sciences Po (France)

---

## Additional Tips

- **Act through your department** — do not email OpenAI directly. Your university's research office is the point of contact
- **The Cal State system inclusion is significant** — it is the largest public university system in the US (23 campuses, 460,000+ students). If you attend any Cal State campus, your institution is technically a partner
- **Research proposals win credits** — universities are more likely to allocate API credits to students with concrete research proposals rather than general exploration
- **Combine with other programs:** Students at partner institutions can also use OpenAI's free ChatGPT tier, the Researcher Access Program (separate), and university-provided compute clusters
- **The consortium may expand** — OpenAI has indicated plans to add more institutions over time
- **Related programs:** OpenAI also offers [ChatGPT Edu](https://openai.com/chatgpt/education/) for universities and the [OpenAI for Startups](https://openai.com/startups/) program for student founders

---

**Sources:**
- [OpenAI — Introducing NextGenAI](https://openai.com/index/introducing-nextgenai/)
- [TechCrunch — OpenAI $50M Grant Program](https://techcrunch.com/2025/03/04/openai-launches-50m-grant-program-to-help-fund-academic-research/)
- [AAU — OpenAI Partners with Research Universities](https://www.aau.edu/newsroom/leading-research-universities-report/openai-partners-aau-members-and-other-research)
- [SiliconANGLE — NextGenAI Consortium Details](https://siliconangle.com/2025/03/04/openai-launches-nextgenai-consortium-15-research-institutions-50m-funding/)
- [University of Oxford — OpenAI Collaboration](https://www.ox.ac.uk/news/2025-03-04-oxford-and-openai-launch-collaboration-advance-research-and-education)
- [University of Michigan — OpenAI Partnership](https://news.umich.edu/u-m-openai-launch-partnership-to-expand-ai-research/)
- [Ohio State — NextGenAI Consortium](https://news.osu.edu/ohio-state-joins-nextgenai-consortium-for-breakthrough-ai-research/)

## Submission 108: NVIDIA NIM Free API Access

- Category: AI API Free Tiers
- Source Link: https://build.nvidia.com/models
- Amount: Free rate-limited API access to 213+ models. Self-hosting on up to 2 nodes / 16 GPUs for dev/test.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

NVIDIA provides free access to **213+ AI models** through their NIM (NVIDIA Inference Microservices) platform at [build.nvidia.com](https://build.nvidia.com). The free tier uses **rate limits** (not credits) for prototyping and development. Available models include DeepSeek V3.2, Llama 3.1, Mistral, Qwen, Gemma, and many more. Members of the free NVIDIA Developer Program can also **self-host NIM containers** on up to 2 nodes / 16 GPUs for dev and testing.

## Registration (Step-by-Step)

1. Go to [build.nvidia.com](https://build.nvidia.com/) and click **Sign In / Sign Up**.
2. Create an NVIDIA account (email + password) or sign in with Google/GitHub.
3. You are automatically enrolled in the **NVIDIA Developer Program** (free).
4. Browse the model catalog and click any model to get an API endpoint and try it in the playground.
5. Generate an **API key** from your account settings to use NIM endpoints programmatically.
6. The API is OpenAI-compatible -- use it as a drop-in replacement in most LLM frameworks.

## Available Models (213+ total)

Key models accessible via free API endpoints:

| Category | Notable Models |
|----------|---------------|
| LLMs | DeepSeek V3.2, DeepSeek R1, Llama 3.1 (8B/70B/405B), Mistral 7B/Large, Qwen 3.5 397B, Kimi K2.5 |
| Code | CodeLlama, StarCoder variants |
| Vision/Multimodal | LLaVA, Phi-3-Vision, NVIDIA VILA |
| Embeddings | NV-Embed, E5-Large |
| Speech | Parakeet, Canary ASR models |
| Image Generation | SDXL, Stable Diffusion variants |
| Specialized | Drug discovery, weather simulation, autonomous vehicles |

## Free Tier Limits

| Resource | Limit |
|----------|-------|
| API access | Rate-limited (varies by model) |
| Rate limit | ~40 RPM (requests per minute) across most models |
| Self-hosting (Developer Program) | Up to 2 nodes / 16 GPUs for dev/test |
| API compatibility | OpenAI-compatible endpoints |
| Commercial use | Prototyping and development only |
| Production use | Requires NVIDIA AI Enterprise license (~$4,500/GPU/year) |

## Additional Tips

- **Credit system was discontinued** in late 2025 and replaced with rate limits. Older references to "1,000 credits" or "5,000 credits" are outdated -- the platform now uses per-model rate limits instead.
- **OpenAI-compatible API** means you can use NIM as a drop-in replacement in LangChain, LlamaIndex, OpenAI SDK, and most other frameworks by just changing the base URL and API key.
- **Self-hosting is the hidden gem:** Developer Program members can download NIM containers and run them on their own hardware (RTX workstations, data centers, or cloud VMs) for free during development.
- **90-day AI Enterprise trial** is available if you need production-grade deployment with support and commercial licensing.
- The **model catalog is massive** (213+ models from 46+ publishers) -- far more variety than most free API tiers.
- Models are optimized with **TensorRT-LLM** for fast inference on NVIDIA hardware.
- Rate limits vary by model -- popular models like DeepSeek and Llama may have tighter limits during peak hours.
- Great for **benchmarking and comparing models** since you can try dozens of LLMs through a single API without managing any infrastructure.

**Sources:**
- https://developer.nvidia.com/nim
- https://build.nvidia.com/models
- https://developer.nvidia.com/blog/access-to-nvidia-nim-now-available-free-to-developer-program-members/
- https://docs.api.nvidia.com/nim/docs/product
- https://forums.developer.nvidia.com/t/request-more-4-000-credits-option-on-build-nvidia-com/344567

## Submission 107: Notion Plus Free for Students & Educators

- Category: Education & Student
- Source Link: https://www.notion.com/product/notion-for-education
- Amount: Free Notion Plus plan (normally $10/month) + 50% off Notion AI add-on (~$5/month instead of $10/month). Student orgs get free Plus with unlimited members.
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Notion offers its **Plus plan completely free** to individual students and teachers at accredited colleges and universities. This is the full Plus plan (normally $10/month) with unlimited pages, blocks, file uploads, and 30-day version history. Students and educators also get **50% off the Notion AI add-on** ($5/month instead of $10/month). Student organizations at verified institutions can also get a free Plus workspace with unlimited members.

---

## Registration (Step-by-Step)

### For Individual Students & Teachers

1. **Create a Notion account** (or log in) using your **educational institution email** (e.g., yourname@university.edu)
2. Go to **Settings & Members** in your Notion workspace
3. Navigate to the **Billing** section (or **Upgrade plan**)
4. Scroll down and click **"Get free Education plan"**
5. If your school is recognized (listed in the World Higher Education Database), approval is automatic
6. If the option does not appear, contact Notion support and request your school's domain to be allowlisted

**Important:** You must use the desktop or web app to upgrade -- mobile app upgrades are not supported for the education plan.

### For Student Organizations

1. Have an official student organization at a verified higher education institution
2. All members must use the same educational email domain
3. Submit an application through Notion's student org application portal
4. Notion's team reviews and contacts applicants with approval status
5. Once approved, the workspace gets a free Plus plan with unlimited members

---

## What You Get

### Plus Plan Features (Free for Edu)

| Feature | Free Plan | Plus Plan (Edu) |
|---|---|---|
| Pages & blocks | Limited | Unlimited |
| File uploads | Up to 5 MB | Unlimited size |
| Version history | 7 days | 30 days |
| Guests | 10 | Up to 100 |
| API access | Yes | Yes |
| Custom domains | No | Yes |
| Custom automations | No | Yes |

### Notion AI Add-on (50% Off)

- Standard price: $10/member/month (or $8/month with annual billing)
- **Education discount: 50% off** -- approximately $5/month
- Notion AI features include: AI writing assistance, Q&A over your workspace, autofill databases, AI summaries, and translation
- The AI add-on is optional and not included in the free Plus plan by default

### Student Organization Workspace

- Free Plus plan for the entire organization
- **Unlimited members** (all must share the same edu email domain)
- Up to 100 guests
- All Plus features included
- Centralized workspace for club documents, projects, and collaboration

---

## Eligibility & Verification

- **Who qualifies:** Students and teachers at accredited colleges and universities
- **K-12 students are NOT eligible** for the individual education plan
- **Verification method:** Educational institution email address only. Notion does not accept student IDs, enrollment letters, or other documentation
- **School must be listed** in the [World Higher Education Database (WHED)](https://www.whed.net/) for automatic recognition
- **One workspace member only** for the individual plan (personal workspace). If you need a team workspace, use the student organization option
- **Duration:** Access continues as long as you are signed in with your educational email. There is no explicit expiration, but access is tied to having a valid edu email
- **If you already have a paid Plus plan:** You must downgrade first before claiming the free education plan

---

## Additional Tips

- **Use your school email from the start.** If you initially signed up with a personal email, you will need to change your account email to your edu address to qualify
- **The free Plus plan is for personal workspaces only** (1 member). For team collaboration, either use the student org option or invite people as guests (up to 100)
- **Notion AI is worth the discounted price** if you write a lot -- the AI writing assistant, workspace Q&A, and auto-fill features can save significant time on essays, research notes, and project management
- **Export your data before graduation.** Once you lose access to your edu email, you may lose the Plus plan. Export important workspaces as Markdown or HTML before your email expires
- **Combine with other student perks:** Notion works well alongside GitHub Student Pack, Figma for Education, and other free student tools for a complete productivity stack
- **Student orgs:** This is one of the few tools that offers a truly free team plan for student clubs. If you run a student organization, this is a no-brainer
- **UNiDAYS discount:** In some regions, Notion also offers discounts through UNiDAYS verification, which may provide additional savings on AI or other add-ons

---

**Sources:**
- [Notion for Education -- Help Center](https://www.notion.com/help/notion-for-education)
- [Notion for Students](https://www.notion.com/product/notion-for-education)
- [Notion Pricing](https://www.notion.com/pricing)
- [Can't Upgrade to Plus Plan for Education -- Help Center](https://www.notion.com/help/cant-upgrade-to-the-plus-plan-for-education)
- [Notion Free for Student Organizations -- Blog](https://www.notion.com/blog/back-to-school-notion-now-free-for-student-organizations)

## Submission 106: Nebius AI Studio Free Credits

- Category: AI API Free Tiers
- Source Link: https://nebius.com/ai-studio
- Amount: Free credits on signup for 60+ models. Startup programs: $5K-$150K.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Nebius AI Studio (now branded as **Token Factory**) gives new users free credits to explore **60+ open-source models** via an **OpenAI-compatible API**. The platform supports text, multimodal, embedding, image generation, and safety models from providers like Meta (Llama), DeepSeek, Qwen, and Mistral. Nebius runs on NVIDIA GPU infrastructure and offers sub-second latency with a 99.9% uptime SLA. For startups, Nebius has multiple programs offering up to **$150K in cloud credits** through partnerships with NVIDIA Inception and other accelerators.

---

## Registration (Step-by-Step)

1. Go to [studio.nebius.ai](https://studio.nebius.ai) (or [tokenfactory.nebius.com](https://tokenfactory.nebius.com))
2. Click **Sign Up** and create your account
3. You will receive free inference credits upon signup
4. Navigate to the API Keys section to generate your API key
5. Set your base URL to the Nebius API endpoint and start making OpenAI-compatible requests
6. Explore the model catalog to choose from 60+ available models

**Important:**
- Free credits are applied automatically to new accounts
- No credit card required for initial free tier
- The API is OpenAI-compatible — change your base URL and API key in existing code
- Zero-retention data mode available for privacy-sensitive workloads

---

## Available AI Models / Services

### Text & Multimodal Models

| Model | Type | Notes |
|---|---|---|
| **DeepSeek R1** | Reasoning | Advanced reasoning model |
| **DeepSeek V3** | Chat | General-purpose chat |
| **Llama 3.3 70B Instruct** | Chat | Meta's versatile instruction-following model |
| **Qwen 2.5 72B** | Chat | Alibaba's large language model |
| **Mistral Nemo Instruct** | Chat | Mistral's efficient model |
| **GPT OSS 120B / 20B** | Chat | Open-source GPT variants |

### Image Generation Models

| Model | Notes |
|---|---|
| **Flux Schnell** | Fast image generation |
| **Flux Dev** | High-quality image generation |
| **SDXL** | Stable Diffusion XL |

### Embedding & Safety Models

| Model | Notes |
|---|---|
| **BAAI/bge-en-icl** | Text embeddings |
| **Qwen3-Embedding-8B** | Multilingual embeddings |
| **Llama Guard 3 8B** | Content safety/moderation |

---

## Free Tier Limits

| Feature | Limit |
|---|---|
| Free credits on signup | Available (amount varies, promotional offers up to $100) |
| Credit card required | No (for free tier) |
| Models available | 60+ open-source models |
| API compatibility | OpenAI-compatible |
| SLA | 99.9% uptime |
| Data retention | Zero-retention mode available |
| Pricing model | Transparent $/token after free credits |

---

## Startup Programs

| Program | Credits | Eligibility |
|---|---|---|
| **AI Explorer** | $5,000 cloud credits | Pre-seed AI startups, no partner required |
| **Nebius for Startups** | Up to $100,000 cloud + $6,000 inference | Startups affiliated with approved partners |
| **AI Lift (NVIDIA Inception)** | Up to $150,000 cloud credits | NVIDIA Inception members |
| **Research Credits** | Varies | Academic researchers |

---

## Additional Tips

- **OpenAI SDK drop-in** — just change the `base_url` to Nebius's endpoint and swap the API key. Works with LangChain, LiteLLM, and other OpenAI-compatible tools
- **Function calling and structured JSON** outputs are natively supported — no extra setup needed
- **RAG-ready** — embedding models like bge-en-icl are available alongside chat models for building retrieval-augmented generation pipelines
- **Fine-tuning** — Nebius supports post-training and fine-tuning workflows on their infrastructure
- **Autoscaling** — the platform handles scaling automatically, you pay per token consumed
- **Speculative decoding** — enabled for faster inference on supported models
- **Earn more credits** — Nebius has periodically offered up to $1,000 for developing GenAI applications on their platform
- **Dedicated endpoints** available for enterprise use with SSO, RBAC, and compliance packages

---

**Sources:**
- [Nebius AI Studio](https://nebius.com/ai-studio)
- [Nebius Token Factory](https://nebius.com/services/token-factory/inference-service)
- [Nebius Studio Portal](https://studio.nebius.ai)
- [Nebius for Startups](https://nebius.com/startup-program)
- [Nebius Research Credits](https://nebius.com/nebius-research-credits-program)
- [Nebius + NVIDIA AI Lift Announcement](https://nebius.com/blog/posts/ai-lift-startups-innovation-with-nvidia)

## Submission 105: Nebius AI Lift Program - Up to $150,000

- Category: Free GPU & Compute
- Source Link: https://nebius.com/startup-program
- Amount: Up to $150,000 cloud credits + $10,000 inference credits (AI Lift); $5,000 free credits (AI Explorer tier)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Nebius AI Cloud's **AI Lift program** offers NVIDIA Inception member startups **up to $150,000 in cloud credits** plus **$10,000 in inference credits**. Announced at NVIDIA GTC 2025, the program provides priority access to H100, H200, and early access to NVIDIA Blackwell (B200/GB200) GPUs. Nebius operates EU-based data centers and claims 90% GPU utilization efficiency. The program includes three standard tiers ($5K / $50K / $100K) plus the enhanced AI Lift tier for Inception members.

---

## Registration (Step-by-Step)

1. **Join NVIDIA Inception first** — go to [nvidia.com/en-us/startups](https://www.nvidia.com/en-us/startups/) and apply. Inception is free, open to all stages, no application fee or deadline
2. Once accepted into NVIDIA Inception, apply for the **Nebius AI Lift** benefits through the Inception portal or directly at [nebius.com/startup-program](https://nebius.com/startup-program)
3. Fill out the Nebius startup application with company details, funding stage, compute needs, and AI product description
4. Nebius reviews your application — they look for venture-backed startups with AI-first products and demonstrated compute demand
5. If approved, credits are applied to your Nebius AI Cloud account with fast-tracked onboarding
6. Start deploying GPU instances, managed Kubernetes clusters, or inference endpoints

**Caveats:**
- The full $150,000 + $10,000 inference package is specifically for **NVIDIA Inception members who join within the first three months** of the program launch
- Rejected applicants can reapply later
- Standard Nebius startup tiers (Explorer/Builder/Scaler) are available even without Inception membership, but with lower credit amounts

---

## Standard Startup Tiers (Without AI Lift)

| Tier | Credits | Requirements |
|---|---|---|
| **AI Explorer** | $5,000 free credits | Under $5M funding, ≤10 employees, AI-first, founded within 5 years, legally incorporated |
| **AI Builder** | Up to $50,000 in discounts | $5M+ raised, AI product beyond prototype, founded within 5 years, approved partner affiliation |
| **AI Scaler** | Up to $100,000 in discounts | $10M+ raised, substantial compute-demanding AI product, founded within 7 years |

## AI Lift Tier (NVIDIA Inception Members)

| Benefit | Detail |
|---|---|
| Cloud credits | Up to $150,000 |
| Inference credits | $10,000 |
| GPU access | H100, H200, early Blackwell (B200/GB200) |
| Technical support | Dedicated AI workload optimization |
| Discounted services | Long-term scaling discounts |
| Co-marketing | Blog features, event speaking opportunities |

---

## GPU Hardware & Specs

Nebius currently offers and is rolling out the following NVIDIA GPUs:

| GPU | VRAM | Memory Type | Approx. Price |
|---|---|---|---|
| **H100 SXM** | 80 GB | HBM3 (3.35 TB/s) | ~$3.00-$4.00/hr |
| **H200 SXM** | 141 GB | HBM3e (4.8 TB/s) | ~$3.50-$7.00/hr |
| **B200** | 192 GB | HBM3e | Pre-order (up to 20% discount) |
| **GB200 NVL72** | Next-gen | HBM3e | Pre-order |
| **GB300 NVL72** | Next-gen | HBM3e | Coming soon |

*Nebius offers up to 35% savings on on-demand rates with long-term commitments.*

---

## Platform Features

- **Managed Kubernetes** and **Slurm-based orchestration** for distributed training
- Fully integrated MLOps stack
- On-demand and reserved GPU options
- High uptime with predictable performance
- Responsive human support (not just chatbots)
- Simple, founder-friendly terms
- EU data center locations (good for GDPR compliance)

---

## Additional Tips

- **NVIDIA Inception is the gateway** — joining Inception is free and fast. It unlocks credits not just from Nebius but from dozens of cloud partners (AWS, Azure, GCP, and many GPU cloud startups)
- **AI Explorer tier is the easiest entry point** — if you don't qualify for AI Lift, the $5,000 Explorer tier has the lowest bar (under $5M funding, ≤10 people)
- **Blackwell early access is a differentiator** — few cloud providers offer B200/GB200 access yet. If you need next-gen hardware for large model training, Nebius is worth considering
- **EU data residency** — if your startup needs to keep data in the EU for regulatory reasons, Nebius is one of the few GPU cloud providers with EU-based infrastructure
- **Compare with other Inception partners** — NVIDIA Inception gives credits across many providers. Stack Nebius credits with credits from other partners to maximize your runway
- **Builder and Scaler tiers offer discounts, not free credits** — only Explorer ($5K) and AI Lift (up to $150K) provide actual free credits. Builder and Scaler provide usage-based discounts

---

**Sources:**
- [Nebius Startup Program](https://nebius.com/startup-program)
- [Nebius AI Lift Announcement Blog](https://nebius.com/blog/posts/ai-lift-startups-innovation-with-nvidia)
- [Nebius GPU Pricing](https://nebius.com/prices)
- [Nebius Blackwell Pre-Order](https://nebius.com/blackwell-pre-order)
- [NVIDIA Inception Program](https://www.nvidia.com/en-us/startups/)
- [TechStartups — Nebius and NVIDIA $150K Credits](https://techstartups.com/2025/04/09/ai-startups-get-a-boost-nebius-and-nvidia-join-forces-to-offer-150k-in-cloud-credits-and-early-blackwell-access/)

## Submission 104: Linode (Akamai) $100 Free Trial Credit

- Category: Cloud Providers
- Source Link: https://www.linode.com/lp/free-credit-100
- Amount: $100 credit for 60 days
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Linode (now part of Akamai) offers **$100 in free credits valid for 60 days** to new customers. The credit covers most Linode services including Shared and Dedicated CPU instances, Kubernetes (LKE), Object Storage, Block Storage, and NodeBalancers. **GPU plans are excluded.** You can spin up to 20 VPS instances, get 20 TB transfer, and access 100+ one-click apps. After the 60 days (or when credits run out), your account auto-converts to pay-as-you-go -- so delete your resources before the trial ends if you don't want charges.

---

## Registration (Step-by-Step)

1. Go to [linode.com/lp/free-credit-100](https://www.linode.com/lp/free-credit-100/) or [akamai.com/lp/free-credit-100](https://www.akamai.com/lp/free-credit-100)
2. Click **Sign Up** and choose your registration method:
   - **Google SSO** -- sign in with your Google account
   - **GitHub SSO** -- sign in with your GitHub account
   - **Email** -- provide your email and create a password
3. Complete your **profile** with name, address, and phone number
4. Add a **valid payment method** for identity verification:
   - Visa, Mastercard, American Express, Discover
   - PayPal or Google Pay
   - Wire transfers and ACH available via Sales
5. **You will not be charged** during the trial unless you exceed the $100 credit
6. Accept the **Terms of Service**
7. $100 credit is applied to your account immediately
8. Start deploying resources from the **Linode Cloud Manager** at [cloud.linode.com](https://cloud.linode.com/)

**Important:** If you use a referral link instead of the official promo page, the credit amount may differ. Use the official $100 credit page for the full amount.

---

## What You Can Use the $100 Credit On

| Service | Description | Starting Price |
|---|---|---|
| **Shared CPU (Essential)** | General purpose VMs | $5/mo (1 GB RAM) |
| **Dedicated CPU** | Compute-intensive workloads | $36/mo (4 GB RAM) |
| **High Memory** | Memory-intensive apps | $60/mo (24 GB RAM) |
| **Linode Kubernetes Engine (LKE)** | Managed Kubernetes (control plane is free) | Worker node pricing |
| **Object Storage** | S3-compatible storage | $5/mo (250 GB) |
| **Block Storage** | Persistent volumes | $0.10/GB/mo |
| **NodeBalancer** | Managed load balancer | $10/mo |
| **Cloud Firewall** | Stateful firewall | Free |
| **Backups** | Automated server backups | From $2/mo |
| **100+ One-Click Apps** | WordPress, GitLab, Docker, etc. | Compute pricing |

**Excluded:** GPU compute plans are not covered by the promotional credit.

---

## Free Tier Limits

| Aspect | Limit |
|---|---|
| **Credit amount** | $100 |
| **Validity period** | 60 days from signup |
| **Max VPS instances** | Up to 20 |
| **Transfer** | Up to 20 TB |
| **GPU access** | Excluded from promo |
| **Overage charges** | Yes -- charged to your payment method |
| **Auto-conversion** | Account becomes pay-as-you-go after trial |
| **Unused credit** | Expires and is forfeited |

---

## Additional Tips

- **Delete resources before day 60** if you don't want to be charged. Your account auto-converts to paid, and running instances will start billing to your credit card
- **$100 goes far on Linode** -- a $5/mo Nanode (1 GB RAM, 1 vCPU, 25 GB SSD) runs for the entire 60-day trial with $90 to spare. You can run ~20 Nanodes simultaneously for the full trial period
- **LKE control plane is free** -- you only pay for worker nodes. Great way to learn Kubernetes on a budget
- **Cloud Firewall is free** -- always enable it, no reason not to
- **One-Click Apps are the fastest way to start** -- deploy WordPress, Discourse, GitLab, or a LAMP stack in minutes
- **Linode is now Akamai Cloud** -- same services, same dashboard, just rebranded. Don't be confused by the dual naming
- **No GPU access on the trial** -- if you need GPU compute for AI/ML, look at other providers (Lambda Labs, Google Colab, Oracle Cloud)
- **Referral program exists** -- after your trial, refer others and earn $100 credit per signup (once they spend $25+)
- **Data centers available in**: Newark, Atlanta, Dallas, Fremont, Toronto, London, Frankfurt, Singapore, Mumbai, Tokyo, Sydney, Jakarta, Chennai, Milan, Miami, Seattle, Chicago, Los Angeles, Washington DC, Paris, Stockholm, Osaka, and Sao Paulo

---

**Sources:**
- [Linode $100 Free Credit](https://www.linode.com/lp/free-credit-100/)
- [Akamai $100 Free Credit](https://www.akamai.com/lp/free-credit-100)
- [Linode Pricing](https://www.linode.com/pricing/)
- [Linode Cloud Manager](https://cloud.linode.com/)
- [ProBlogGuru: Linode Free Trial](https://www.problogguru.com/linode-free-trials/)

## Submission 103: Leonardo AI $5 Free API Credits

- Category: AI API Free Tiers
- Source Link: https://leonardo.ai/api
- Amount: $5 free API credits (one-time) + 150 fast tokens/day on main platform
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Leonardo AI gives every new API account **$5 in free credits** for AI image and video generation through its API. No credit card required to claim. The main platform also offers **150 free daily tokens** (non-API usage) on the Free plan, enough for ~6-12 image generations per day depending on settings. Leonardo is best known for its flagship **Phoenix** model (prompt-adherent, photorealistic) and **Kino XL** (cinematic style). The API is pay-as-you-go after the initial $5, with no monthly minimums.

---

## Registration (Step-by-Step)

1. Go to [leonardo.ai](https://leonardo.ai) and click **"Launch App"** or sign up
2. Create an account using **email**, **Google**, **Apple**, or **Microsoft** login
3. Once logged in, navigate to the **API Access** section from the left sidebar or go to [app.leonardo.ai/api-access](https://app.leonardo.ai/api-access)
4. Click **"Create New Key"** and give your API key a descriptive name (e.g., `my-app-prod`)
5. Copy your API key immediately — it will not be shown again
6. You automatically receive **$5 in free API credits** upon creating your first API key
7. Start making requests to the Leonardo API endpoints

**Important:**
- Keep your API key secure — do not embed it in client-side code
- The $5 credit is one-time and does not renew
- Free platform tokens (150/day) are separate from API credits
- All generations on the free plan are **public** and visible in the community feed

---

## Available AI Models / Services

### Image Generation Models

| Model | Type | Notes |
|---|---|---|
| **Phoenix** | Foundational (in-house) | Flagship model; high prompt adherence, photorealistic output |
| **Kino XL** | Cinematic | Excels in wider aspect ratios, dramatic compositions, no negative prompts needed |
| **FLUX Dev** | Community | High-quality text-to-image |
| **FLUX Schnell** | Community | Fast generation variant |
| **FLUX.2 Pro** | Community | Advanced FLUX variant |
| **GPT Image-1.5** | OpenAI | Image generation via OpenAI model |
| **Ideogram 3.0** | Third-party | Text rendering specialist |

### Video Generation Models

| Model | Notes |
|---|---|
| **Kling 2.1 Pro** | High-quality video generation |
| **Veo 3.0 / 3.1** | Google's video models |
| **Seedance** | Multiple configuration options |

---

## Free Tier Limits

### Platform (Non-API) Free Plan

| Feature | Limit |
|---|---|
| Daily tokens | 150 (no rollover) |
| Monthly equivalent | ~4,500 tokens |
| Standard 512x512 image | ~12 tokens |
| HD 1024x1024 image | ~24 tokens |
| PhotoReal v3 image | ~30 tokens |
| Generation visibility | Public only |
| Queue priority | Low (5-20 min wait) |

### API Free Credits

| Feature | Limit |
|---|---|
| Initial free credits | $5 one-time |
| Credit card required | No |
| Credit expiration | Until used |
| Billing model | Pay-as-you-go after $5 |

### Paid Platform Plans (for reference)

| Plan | Price/Month | Daily Tokens |
|---|---|---|
| Starter | $15 ($12/yr) | 8,500 |
| Creator | $35 ($28/yr) | 25,000 |
| Professional | $70 ($56/yr) | 60,000 |
| Enterprise | ~$300+ | Custom |

---

## Additional Tips

- **Use the Pricing Calculator** at [app.leonardo.ai/api-access/pricing-calculator](https://app.leonardo.ai/api-access/pricing-calculator) to estimate how many API credits different generation configurations consume before spending your free $5
- **Token Bank** on paid plans allows limited rollover (25K-180K depending on tier) — free plan tokens do not roll over
- **Video generation** consumes significantly more tokens/credits than images — be careful with your free $5
- **Private generations** require a paid plan — everything on the free tier is visible to the community
- **Phoenix model** consistently ranks high in prompt adherence benchmarks — it is the best starting point for most use cases
- **API top-ups** vary in price by plan tier — if you plan to use the API heavily, consider a subscription for better per-credit rates
- Leonardo has a Python community library: [github.com/wwakabobik/leonardo_api](https://github.com/wwakabobik/leonardo_api)

---

**Sources:**
- [Leonardo AI API Page](https://leonardo.ai/api)
- [Leonardo AI Pricing](https://leonardo.ai/pricing)
- [Leonardo AI Documentation](https://docs.leonardo.ai)
- [Leonardo AI Pricing Calculator](https://docs.leonardo.ai/docs/plan-with-the-pricing-calculator)
- [Leonardo AI Pricing FAQs](https://docs.leonardo.ai/docs/pricing-and-plans-faq)
- [Leonardo AI Phoenix Model](https://leonardo.ai/phoenix/)

## Submission 102: Koyeb Serverless Free Tier

- Category: Cloud Providers
- Source Link: https://www.koyeb.com/pricing
- Amount: 1 free web service (512 MB RAM, 0.1 vCPU) + 1 free Postgres database (1 GB, 5 hrs/month) + 5 custom domains
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Koyeb offers a forever-free Starter plan with one free web service and one free PostgreSQL database. No credit card required. Unlike Render's free tier, Koyeb's free web service never expires and you will never be charged for it. Apps support scale-to-zero. Good for deploying small APIs, bots, and side projects that need to stay running without a monthly bill.

## Registration (Step-by-Step)

1. Go to [koyeb.com](https://www.koyeb.com/) and sign up with GitHub or email.
2. No credit card required for the Starter plan.
3. Create a new **Service** from the dashboard -- deploy from a Git repository, Docker image, or pre-built template.
4. Your free web service runs on a **Nano instance** (512 MB RAM, 0.1 vCPU, 2 GB SSD).
5. Optionally create a free PostgreSQL database (limited to 5 active hours/month and 1 GB storage).
6. Attach up to 5 custom domains at no cost.

## Free Tier Limits

| Resource | Free Tier |
|----------|----------|
| Web Services | 1 free Nano instance (512 MB RAM, 0.1 vCPU, 2 GB SSD) |
| PostgreSQL Database | 1 free database (1 GB storage, 5 active hours/month) |
| Custom Domains | 5 included |
| Regions | Frankfurt (EU) or Washington D.C. (US) only |
| Bandwidth | 1 TB/month included (overage: $0.02/GB US/EU, $0.04/GB Asia) |
| Team Members | 1 user on Starter plan |
| Scale-to-Zero | Enforced on free instances |
| Billing | Never charged for the free web service |

## Additional Tips

- **Scale-to-zero is enforced** on free instances, meaning your service scales down when idle and scales back up on incoming requests. Expect cold starts.
- **5 active hours/month on the free database** is very limited -- it is best for lightweight usage like dev/test environments, not production databases with constant traffic.
- **The free web service is permanent** -- unlike Render, it does not expire. Koyeb has publicly committed to maintaining a forever-free tier.
- Deploy from **Git, Docker, or pre-built templates** -- supports Node.js, Python, Go, Rust, and more.
- **Horizontal scaling** is natively supported even on the free tier, though you only get 1 free instance.
- To avoid unexpected charges, be careful when adding services beyond the free one -- paid instances on the Starter plan use pay-as-you-go pricing (billed per second).
- For more compute, the Starter plan also includes $1/month in compute credit toward paid instances.
- Koyeb supports **GPU instances** on paid plans if you need to scale up later (A100, L4, T4 available).

**Sources:**
- https://www.koyeb.com/pricing
- https://www.koyeb.com/docs/faqs/pricing
- https://www.koyeb.com/blog/sustaining-free-compute-in-a-hostile-environment
- https://getdeploying.com/koyeb

## Submission 101: Kluster.ai Free Batch Inference Tier

- Category: AI API Free Tiers
- Source Link: https://www.kluster.ai/
- Amount: Free tier: up to 1,000 batch requests/file, 100MB max file size. Promotional credits periodically available.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Kluster.ai is an AI cloud platform specializing in **batch inference** for open-source models like **DeepSeek-R1**, **Llama 4 Maverick/Scout**, and **Qwen3**. The free tier supports up to **1,000 batch requests per file** with a 100MB max file size. The platform also offers adaptive real-time inference with sub-second latency. Kluster.ai uses an **OpenAI-compatible API** (`api.kluster.ai/v1`), so you can use the standard OpenAI Python SDK. Batch inference pricing starts as low as **$0.10/M tokens** for smaller models with 72-hour completion windows — up to 50% cheaper than competitors.

---

## Registration (Step-by-Step)

1. Go to [kluster.ai](https://www.kluster.ai) and click **Sign Up**
2. Create your account
3. Navigate to the API section to generate your API key
4. Set your base URL to `https://api.kluster.ai/v1`
5. Use the OpenAI Python SDK with your kluster.ai API key
6. Prepare your batch request as a **JSONL file** (one request per line)
7. Submit your batch job via the API

**Important:**
- Free tier has a hard limit of 1,000 requests per batch file
- Max file size is 100MB per batch file (applies to all tiers)
- The API is OpenAI-compatible — use `from openai import OpenAI` and change the base URL
- Batch jobs are processed asynchronously — you submit and poll for results

---

## Available AI Models

| Model | Category | Notes |
|---|---|---|
| **DeepSeek-R1** | Reasoning | 671B parameter reasoning model |
| **Llama 4 Maverick** | Chat | Meta's latest MoE model |
| **Llama 4 Scout** | Chat | Meta's efficient MoE model |
| **Llama 3.3 70B Instruct** | Chat | Meta's instruction-following model |
| **Llama 3.1 405B Instruct Turbo** | Chat | Meta's largest Llama model |
| **Llama 3.1 8B** | Chat | Fast, lightweight model |
| **Qwen3-235B-A22B** | Chat | Alibaba's flagship MoE |
| **Gemma 3** | Chat | Google's open model |

---

## Free Tier Limits

| Feature | Free Tier | Standard Tier |
|---|---|---|
| Max batch requests per file | 1,000 | Unlimited |
| Max file size | 100MB | 100MB |
| Models available | All supported | All supported |
| Completion windows | 24h, 48h, 72h | 24h, 48h, 72h |
| Real-time inference | Limited | Yes |

---

## Batch Inference Pricing (per 1M tokens)

| Model | 24-Hour | 48-Hour | 72-Hour |
|---|---|---|---|
| **DeepSeek-R1** | $3.50 | $3.00 | $2.50 |
| **Llama 4 Scout 17Bx16E** | $0.15 | $0.12 | $0.10 |
| **Llama 3.1 405B Turbo** | Higher | Mid | Lower |

*Longer completion windows = lower cost. Choose 72h for maximum savings on non-urgent workloads.*

---

## Inference Modes

| Mode | Latency | Best For |
|---|---|---|
| **Real-time** | Sub-second | Interactive apps, chat |
| **Asynchronous** | Minutes | Flexible timing, moderate volume |
| **Batch** | 24-72 hours | High-volume, bulk processing |

---

## Additional Tips

- **Batch inference is the killer feature** — if you have large-scale processing jobs (data labeling, content generation, analysis), batch mode at 72h completion can be 50%+ cheaper than real-time
- **JSONL format** — each line in your batch file is a separate request in JSON format. Make sure your file is valid JSONL before submitting
- **Use with Bespoke Curator** — kluster.ai integrates with Bespoke Labs' data curation tool for efficient large-scale inference pipelines
- **OpenAI SDK compatible** — no custom SDK needed. Just `pip install openai`, set the base URL to `api.kluster.ai/v1`, and use your API key
- **Free tier is per-file, not per-month** — you can submit multiple batch files with up to 1,000 requests each
- **Promotional credits** — kluster.ai has periodically offered $100 in free credits (e.g., for DeepSeek-R1 usage). Watch their blog for announcements
- **Adaptive inference** — the platform dynamically adjusts computing resources based on workload, which helps keep costs low
- **Compare with alternatives** — for batch inference, also consider OpenAI Batch API (50% discount), Together.ai, and Fireworks.ai

---

**Sources:**
- [Kluster.ai](https://www.kluster.ai)
- [Kluster.ai Documentation](https://docs.kluster.ai)
- [Kluster.ai Supported Models](https://docs.kluster.ai/get-started/models/)
- [Kluster.ai Batch Inference Guide](https://docs.kluster.ai/get-started/start-building/batch/)
- [Kluster.ai Adaptive Inference Blog](https://www.kluster.ai/blog/introducing-kluster-ai-s-adaptive-inference-scalable-on-demand-access-to-llama-3-3-3-1-models)
- [Kluster.ai on Artificial Analysis](https://artificialanalysis.ai/providers/klusterai)
- [Using kluster.ai with Bespoke Curator](https://docs.bespokelabs.ai/bespoke-curator/save-usdusdusd-on-llm-inference/using-kluster.ai-for-batch-inference)

## Submission 100: Hyperbolic Free Tier AI Inference

- Category: AI API Free Tiers
- Source Link: https://www.hyperbolic.ai/
- Amount: Basic tier: free (60 RPM) + $1 promotional credit. Pro tier ($5 deposit): 600 RPM. 25+ models at up to 80% less cost.
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Hyperbolic is an **open-access AI cloud** offering serverless inference at up to **80% less cost** than traditional providers. The Basic tier starts **free** with a **$1 promotional credit** (phone verification required) and **60 RPM** rate limits. The platform supports **25+ models** across text, image, audio, and video via an **OpenAI-compatible API** with zero data retention. Access models like **Llama 3.1 405B**, **DeepSeek V3**, **Qwen 2.5 72B**, and **GPT OSS 120B** at extremely competitive per-token pricing. Upgrade to Pro ($5 deposit) for **600 RPM** and higher throughput.

---

## Registration (Step-by-Step)

1. Go to [hyperbolic.ai](https://www.hyperbolic.ai) and click **Sign Up**
2. Create an account and complete **phone verification**
3. You receive a **$1 promotional credit** after verification
4. Navigate to [app.hyperbolic.ai/models](https://app.hyperbolic.ai/models) to browse available models
5. Generate your API key from the dashboard
6. Set the base URL to Hyperbolic's OpenAI-compatible endpoint
7. Start making requests — you are on the **Basic tier** (60 RPM)

**Important:**
- The $1 promotional credit **cannot be used for GPU rentals** — only for serverless inference
- Phone verification is required to claim the $1 credit
- Basic tier is free to use within rate limits
- The API is OpenAI-compatible — change `base_url` and API key in existing code
- Zero data retention policy — your prompts are not stored

---

## Available AI Models

### Text/LLM Models (per 1M tokens)

| Model | Price (Input/Output) | Notes |
|---|---|---|
| **GPT OSS 120B (FP8)** | $0.30 | Open-source GPT variant |
| **GPT OSS 20B** | Lower | Lightweight GPT variant |
| **DeepSeek V3** | $0.25 | General-purpose chat |
| **DeepSeek R1** | Varies | Reasoning model |
| **Qwen 2.5 72B Instruct** | $0.40 | Multilingual, large context |
| **Llama 3.1 405B Instruct** | $4.00 | Meta's largest open model |
| **Llama 3.1 70B / 8B** | Lower | Versatile Llama variants |

### Image Generation

| Model | Base Price | Scaling |
|---|---|---|
| **Flux.1** | $0.01 | Scales by (width/1024) x (height/1024) x (steps/25) |
| **Stable Diffusion** | $0.01 | Same scaling formula |

### Audio

| Model | Price |
|---|---|
| **Melo TTS** | $5.00 per 1M characters |

---

## Tier Comparison

| Feature | Basic (Free) | Pro ($5 deposit) | Enterprise (Custom) |
|---|---|---|---|
| Rate limit | 60 RPM | 600 RPM | Unlimited |
| Promotional credit | $1 | — | — |
| Deposit required | No (phone verification) | $5 minimum | Custom |
| Text/Image/Audio/Video | Yes | Yes | Yes |
| Fine-tuning | Yes | Yes | Yes |
| Dedicated instances | No | No | Yes |
| Custom SLAs | No | No | Yes |
| Dedicated support | No | No | Yes |

### Special Rate Limits (Basic vs Pro)

| Model | Basic RPM | Pro RPM |
|---|---|---|
| **Standard models** | 60 | 600 |
| **Llama 3.1 405B** | 5 | 120 |
| **Flux.1** | 1 per 5 min | 50 |

---

## Enterprise GPU Rental Pricing

| GPU | Price/Hour | Notes |
|---|---|---|
| **RTX 3090** | $0.30 | Budget option |
| **RTX 4090** | $0.50 | Consumer-grade |
| **A100 PCIe** | $1.60 | Data center standard |
| **A100 SXM** | $1.80 | Higher bandwidth |
| **H100 PCIe** | $3.00 | Latest generation |
| **H100 SXM** | $3.20 | Top-tier performance |

*Compared to AWS H100 at $3.90/hour, Hyperbolic offers 62% savings.*

---

## Additional Tips

- **The $1 credit goes further than you think** — with DeepSeek V3 at $0.25/M tokens, that is 4 million tokens of inference. Enough for significant prototyping
- **Phone verification is mandatory** — you cannot get the promotional credit without it
- **Basic tier 405B limits are tight** — only 5 RPM for Llama 3.1 405B on Basic. Use smaller models for rapid iteration, then switch to 405B for final outputs
- **Flux.1 image gen is extremely cheap** — at $0.01 base price, you can generate 100 standard images for $1
- **Upgrade to Pro is painless** — deposit $5 to get 10x rate limits across all models. The $5 becomes usable credit, not a fee
- **Zero data retention** — Hyperbolic does not store your prompts or outputs, making it suitable for sensitive workloads
- **OpenAI SDK compatible** — use `pip install openai`, set the base URL to Hyperbolic's endpoint, and swap your API key
- **GPU rentals are separate** — the $1 promotional credit is only for serverless inference, not GPU rentals
- **Compare with alternatives** — for cheap inference, also check SiliconFlow, Together.ai, and Fireworks.ai. For privacy-first, compare with Venice AI

---

**Sources:**
- [Hyperbolic AI](https://www.hyperbolic.ai)
- [Hyperbolic Pricing Documentation](https://docs.hyperbolic.xyz/docs/hyperbolic-pricing)
- [Hyperbolic Models Catalog](https://app.hyperbolic.ai/models)
- [Hyperbolic on Artificial Analysis](https://artificialanalysis.ai/providers/hyperbolic)
- [Hyperbolic on Hugging Face](https://huggingface.co/docs/inference-providers/en/providers/hyperbolic)
- [Hyperbolic Stable Diffusion Pricing Blog](https://www.hyperbolic.ai/blog/stable-diffusion-pricing)

## Submission 99: Hugging Face ZeroGPU for Spaces - Free H200 GPU Access

- Category: Free GPU & Compute
- Source Link: https://huggingface.co/docs/hub/en/spaces-zerogpu
- Amount: Free GPU allocation for Gradio Spaces (NVIDIA H200, 70-141GB VRAM). Free accounts: 3.5 min/day. PRO: 25 min/day.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

ZeroGPU is Hugging Face's **dynamic GPU sharing system** that allocates NVIDIA H200 GPUs to Gradio-based Spaces on demand. Instead of renting a dedicated GPU 24/7, your app grabs a powerful H200 slice when it needs compute and releases it immediately after. Free accounts get **3.5 minutes of daily GPU quota**, and PRO subscribers ($9/month) get **25 minutes/day with highest queue priority**. All users can use existing ZeroGPU Spaces for free; hosting your own requires a PRO or Team/Enterprise subscription.

---

## Registration (Step-by-Step)

### Using Existing ZeroGPU Spaces (Free)
1. Go to [huggingface.co](https://huggingface.co) and create a free account
2. Browse the [curated list of ZeroGPU Spaces](https://huggingface.co/spaces/enzostvs/zero-gpu-spaces)
3. Use any Space directly -- GPU is allocated automatically when you run inference
4. Even unauthenticated users get 2 minutes/day, but logging in bumps you to 3.5 minutes

### Hosting Your Own ZeroGPU Space (PRO Required)
1. Subscribe to [Hugging Face PRO](https://huggingface.co/subscribe/pro) ($9/month) or join a Team/Enterprise org
2. Create a new Space and select **Gradio** as the SDK
3. In Space settings, choose **ZeroGPU** as the hardware
4. In your code, `import spaces` and decorate GPU-dependent functions with `@spaces.GPU`
5. Push your code -- the Space auto-builds and GPU is allocated dynamically per request

**Important:**
- ZeroGPU **only works with Gradio SDK** -- Streamlit and Docker Spaces are not supported
- Personal PRO accounts can host up to **10 ZeroGPU Spaces**
- Organization Team/Enterprise accounts can host up to **50 ZeroGPU Spaces**
- `torch.compile` is not supported; use PyTorch ahead-of-time compilation (torch 2.8+) instead

---

## GPU Hardware & Specs

| GPU Size | Backing Hardware | VRAM | Quota Cost |
|---|---|---|---|
| **large** (default) | Half NVIDIA H200 | 70 GB | 1x |
| **xlarge** | Full NVIDIA H200 | 141 GB | 2x |

The NVIDIA H200 features HBM3e memory and Hopper architecture, making it one of the most powerful GPUs available. You get access to this hardware for free -- a significant advantage over platforms offering older T4 or P100 GPUs.

### GPU Size Selection in Code

```python
# Default: half H200 (70GB VRAM)
@spaces.GPU
def generate(prompt):
    return pipe(prompt).images

# Full H200 (141GB VRAM) -- consumes 2x quota
@spaces.GPU(size="xlarge")
def generate_large(prompt):
    return pipe(prompt).images
```

---

## Free Tier Limits

| Account Type | Daily GPU Quota | Queue Priority |
|---|---|---|
| **Unauthenticated** | 2 minutes | Low |
| **Free account** | 3.5 minutes | Medium |
| **PRO account** ($9/month) | 25 minutes | Highest |
| **Team org member** | 25 minutes | Highest |
| **Enterprise org member** | 45 minutes | Highest |

| Limit | Value |
|---|---|
| **Quota reset** | 24 hours after first GPU usage |
| **Default function timeout** | 60 seconds |
| **Max function duration** | Configurable via `@spaces.GPU(duration=120)` |
| **ZeroGPU Spaces limit (PRO)** | 10 per personal account |
| **ZeroGPU Spaces limit (Org)** | 50 per organization |

---

## Supported Frameworks & Versions

| Framework | Supported Versions |
|---|---|
| **Gradio** | 4+ (required -- only supported SDK) |
| **PyTorch** | 2.1.0 through latest (2.9.1) |
| **Python** | 3.10.13, 3.12.12 |
| **HF Libraries** | transformers, diffusers, accelerate (enhanced compatibility) |

---

## Additional Tips

- **Maximize quota efficiency**: Set shorter `duration` values for quick functions to improve queue priority. Example: `@spaces.GPU(duration=30)` for functions that finish in under 30 seconds
- **Use dynamic duration**: Pass a callable to `duration` that estimates runtime based on input parameters, so longer tasks get appropriate time without wasting quota on short ones
- **xlarge is rarely needed**: Only use `size="xlarge"` when your model truly needs >70 GB VRAM. It consumes 2x quota and has longer queue wait times
- **Ahead-of-time compilation**: Since `torch.compile` is not supported, use PyTorch AOT compilation (torch 2.8+) for performance optimization. See the [official blog post](https://huggingface.co/blog/zerogpu-aoti)
- **Use flash-attention 3**: For large models, combining AOT compilation with flash-attention 3 significantly reduces inference time and quota consumption
- **The `@spaces.GPU` decorator is effect-free outside ZeroGPU**: Your code will work normally on local machines or other platforms without modification
- **Remaining quota affects priority**: The more quota you have left, the higher your priority in the queue. Heavy users get deprioritized within a day
- **PRO subscription is excellent value**: At $9/month, you get 25 minutes of H200 daily -- equivalent to hundreds of dollars of GPU time per month if rented conventionally

---

**Sources:**
- [Spaces ZeroGPU Documentation](https://huggingface.co/docs/hub/en/spaces-zerogpu)
- [ZeroGPU AOT Compilation Blog Post](https://huggingface.co/blog/zerogpu-aoti)
- [Hugging Face Pricing](https://huggingface.co/pricing)
- [Hugging Face PRO Account](https://huggingface.co/pro)
- [ZeroGPU Explorers Community](https://huggingface.co/zero-gpu-explorers)
- [Advanced Compute Options](https://huggingface.co/docs/hub/advanced-compute-options)
- [Using GPU Spaces](https://huggingface.co/docs/hub/en/spaces-gpus)

## Submission 98: Hugging Face Classrooms (Free ML Platform for Education)

- Category: Education & Student
- Source Link: https://huggingface.co/classrooms
- Amount: Free classroom workspace + dataset/model hosting + Accelerated Inference API + free courses and educator toolkit (8 languages)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Hugging Face Classrooms gives teachers and students **free collaborative workspaces** for machine learning on the Hugging Face Hub. Instructors create a classroom, invite students, and everyone can upload datasets, train models, build demos, and deploy -- all hosted for free. Classrooms also get **free access to the Accelerated Inference API**, which provides advanced computational resources beyond what standard free accounts receive. Combined with Hugging Face's free courses (NLP, Deep RL, Transformers) and an educator toolkit translated into 8 languages, this is one of the most comprehensive free ML education platforms available.

---

## Registration (Step-by-Step)

1. **Create a free Hugging Face account** at [huggingface.co](https://huggingface.co/) if you do not already have one
2. **Go to the Classrooms page:** [huggingface.co/classrooms](https://huggingface.co/classrooms)
3. **Click "Create a classroom"** -- this creates a new organization on the Hub with classroom-specific features
4. **Configure your classroom:**
   - Set the classroom name and description
   - Choose visibility (private or public)
   - Set member permissions (admin, read, write)
5. **Invite students** using their Hugging Face usernames or email addresses
6. Students join the classroom and can immediately start uploading datasets, models, and demos to the shared workspace
7. **Optional:** Review the [Getting Started Guide](https://huggingface2.notion.site/How-to-get-started-with-Hugging-Face-for-Classrooms-045b0ab2a7a646f18dfaef772cf146d9) for detailed setup instructions

No institutional verification or application process is required -- any educator can create a classroom.

---

## What You Get

### Classroom Workspace

- **Shared Hub organization** -- a centralized space where all classroom members can collaborate
- **Model hosting** -- upload, version-control, and share ML models for free
- **Dataset hosting** -- upload and manage datasets within the classroom
- **Spaces (ML demos)** -- build and deploy interactive ML demos using Gradio or Streamlit, hosted free on Hugging Face
- **Version control** -- Git-based versioning for all models, datasets, and code
- **Access controls** -- admin, read, and write permissions for students
- **Private or public** -- choose whether your classroom content is visible to the broader community

### Computational Resources

- **Accelerated Inference API** -- classrooms get free access to advanced inference capabilities, allowing students to run models without needing their own GPU hardware
- **Hugging Face Spaces** -- free hosting for Gradio and Streamlit ML demos (CPU-based by default; GPU Spaces available at cost)
- **Google Colab integration** -- many course materials include ready-to-run Colab notebooks with free GPU access

### Free Educational Resources

| Resource | Description |
|---|---|
| **NLP Course** | Full course on Transformers for NLP, free, no ads |
| **Deep RL Course** | Theory + practice for reinforcement learning |
| **ML Demos Course** | Learn to build interactive ML interfaces with Gradio |
| **Educator Toolkit** | Lab/homework/class materials, translated into 8 languages |
| **Transformers Book** | Comprehensive guide on the Transformers library |
| **GitHub Education Toolkit** | [github.com/huggingface/education-toolkit](https://github.com/huggingface/education-toolkit) |

---

## Eligibility & Verification

- **No formal verification required** -- any educator or teacher can create a classroom for free
- **No institutional email required** -- unlike many education programs, Hugging Face does not gate access behind .edu emails
- **Students need a free Hugging Face account** -- no paid plan required
- **No limit on classroom size** -- add as many students as needed
- **Self-service setup** -- no application, approval process, or waiting period

---

## Additional Tips

- **Best for hands-on ML courses** -- Classrooms shine when students are actively building, training, and deploying models. If your course is purely theoretical, you may not need the full classroom setup
- **Pair with Google Colab** -- Hugging Face's course notebooks are designed to run in Google Colab, giving students free GPU access for training. This combination (HF Classrooms + Colab) is a powerful zero-cost ML education stack
- **Use Spaces for student projects** -- Students can deploy their final projects as interactive Gradio or Streamlit demos on Hugging Face Spaces, creating a shareable portfolio piece
- **The Educator Toolkit is underrated** -- pre-built lab materials in 8 languages save significant course prep time. Check the [GitHub repo](https://github.com/huggingface/education-toolkit) for the latest materials
- **Request expert support** -- Hugging Face offers to send team members to conduct sessions for your class. Reach out through their education page if interested
- **For individual learners** -- you do not need to be in a formal classroom to benefit. All courses, models, and datasets on Hugging Face are freely accessible to anyone with an account
- **Compare with Kaggle** -- Kaggle offers free GPU notebooks but lacks the collaborative classroom workspace. Hugging Face Classrooms is better for structured courses; Kaggle is better for competitions and individual exploration
- **Student Pro Plan** -- Hugging Face also offers a Pro plan ($9/month) with additional GPU resources. As of early 2026, there is no official student discount on Pro, but check the [Hugging Face Forums](https://discuss.huggingface.co/) for any emerging education pricing

---

**Sources:**
- [Hugging Face Classrooms](https://huggingface.co/classrooms)
- [Introducing Hugging Face for Education -- Blog](https://huggingface.co/blog/education)
- [Hugging Face Education Toolkit -- GitHub](https://github.com/huggingface/education-toolkit)
- [Hugging Face Classrooms -- OECD.AI](https://oecd.ai/en/catalogue/tools/hugging-face-classrooms)
- [Hugging Face Learn](https://huggingface.co/learn)

## Submission 97: Hetzner Cloud Free Credits - EUR 20

- Category: Cloud Providers
- Source Link: https://www.hetzner.com/cloud
- Amount: EUR 20 free credits (referral) + up to EUR 50 per published community tutorial
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Hetzner Cloud offers **EUR 20 in free credits** for new customers who sign up through a referral link. Credits are valid until the end of the month following activation. Hetzner is known for aggressive pricing -- their cheapest VPS starts at EUR 3.49/mo, so EUR 20 gets you several months of small-instance hosting. Additionally, you can earn **up to EUR 50 in credits** by writing and publishing tutorials on the Hetzner Community platform. Hetzner operates its own data centers in Germany and Finland, with additional locations in the US (Ashburn, Hillsboro) and Singapore.

---

## Registration (Step-by-Step)

### Getting the EUR 20 Referral Credit

1. Find a **Hetzner referral link** -- search for "Hetzner referral" or ask a friend who uses Hetzner. You must use a referral link to get the EUR 20 credit
2. Click the referral link to go to Hetzner Cloud signup
3. Go to [console.hetzner.cloud](https://console.hetzner.cloud/) and create an account
4. Enter your **name, email, and address**
5. Verify your email address
6. Add a **valid payment method** -- credit card, PayPal, or SEPA direct debit (EU)
7. Complete identity verification if prompted
8. **EUR 20 credit** is applied to your Cloud account automatically
9. Start deploying servers from the Hetzner Cloud Console

**Important:** You must not have been a Hetzner customer within the past 12 months to qualify for the referral credit.

### Getting the EUR 50 Tutorial Credit

1. Go to the [Hetzner Community](https://community.hetzner.com/tutorials/) and review existing tutorials
2. Read the [Contribution Guidelines](https://github.com/hetzneronline/community-content/blob/master/contributing.md) on GitHub
3. Write an original tutorial about free/open-source software (development, sysadmin, web technology)
4. Open a **Pull Request** on the [hetzneronline/community-content](https://github.com/hetzneronline/community-content) GitHub repository
5. A Community Manager reviews your tutorial and provides feedback
6. Once published, you receive **up to EUR 50** in Hetzner Cloud credits (amount depends on tutorial depth and quality)
7. Major updates to existing tutorials earn approximately **EUR 10** in credits

**Restrictions on tutorials:**
- Must be original content -- no republishing from other sources
- Must not be primarily or entirely AI-generated
- Must cover free and open-source software topics

---

## Hetzner Cloud Services & Pricing

| Service | Specs | Starting Price |
|---|---|---|
| **Shared (Cost-Optimized)** | 1 vCPU, 2 GB RAM, 20 GB SSD | EUR 3.49/mo |
| **Shared (Regular)** | 2 vCPU, 2 GB RAM, 40 GB SSD | EUR 4.99/mo |
| **Dedicated General Purpose** | 2 vCPU, 8 GB RAM, 80 GB SSD | EUR 12.49/mo |
| **Load Balancers** | Managed LB | EUR 6.49/mo |
| **Block Storage** | Persistent volumes | EUR 0.052/GB/mo |
| **Snapshots** | Server images | EUR 0.012/GB/mo |
| **Floating IPs** | Static IPs | EUR 3.49/mo |
| **Private Networks** | VLANs | Free |
| **Firewalls** | Stateful firewalls | Free |
| **DDoS Protection** | Included | Free |

**Billing model:** Hybrid hourly/monthly. Your bill never exceeds the monthly cap. If you delete a server mid-month, you only pay for the hours used.

---

## Free Tier Limits

| Aspect | Limit |
|---|---|
| **Referral credit** | EUR 20 |
| **Tutorial credit** | Up to EUR 50 per published tutorial |
| **Referral credit validity** | End of the month following activation |
| **Referral credit eligibility** | Not a Hetzner customer in past 12 months |
| **Referrer bonus** | EUR 10 per referred customer (after they spend EUR 10+) |
| **Credit applicability** | Hetzner Cloud products only |
| **Cash-out** | Credits cannot be exchanged for cash |

---

## Available Regions

| Location | Type | Data Center |
|---|---|---|
| **Nuremberg, Germany** | Own DC | FSN1 |
| **Falkenstein, Germany** | Own DC | NBG1 |
| **Helsinki, Finland** | Own DC | HEL1 |
| **Ashburn, Virginia, USA** | Third-party DC | ASH1 |
| **Hillsboro, Oregon, USA** | Third-party DC | HIL1 |
| **Singapore** | Third-party DC | SIN1 |

---

## Additional Tips

- **EUR 20 goes surprisingly far** -- Hetzner's cheapest VPS (EUR 3.49/mo) means your credit covers ~5.7 months of a small instance. That's one of the best credit-to-compute ratios of any provider
- **Credit validity is short** -- EUR 20 expires at the end of the month following activation. If you sign up on Feb 1, credit expires March 31. If you sign up Feb 28, it still expires March 31. Sign up early in the month to maximize usage time
- **No free tier without referral** -- unlike AWS or Oracle, Hetzner doesn't offer free credits without a referral link. You need one to get the EUR 20
- **Tutorial credits stack** -- you can write multiple tutorials and earn EUR 50 each. This is a legitimate way to get ongoing free hosting if you enjoy writing technical content
- **AI-generated tutorials are rejected** -- Hetzner explicitly bans AI-written submissions. Write your tutorials yourself
- **Hetzner's pricing is already dirt cheap** -- even after credits run out, their VPS pricing significantly undercuts AWS, GCP, and Azure. A 2 vCPU / 4 GB RAM instance is under EUR 6/mo
- **GDPR-friendly** -- Hetzner's European data centers and German company registration make it a strong choice for EU-based projects with data residency requirements
- **REST API and CLI available** -- full API for automation, Terraform provider exists
- **Referral abuse is prohibited** -- don't create multiple accounts or use your own referral link. Hetzner monitors for abuse and will revoke credits
- **24/7 support included** for all customers, even during the free credit period

---

**Sources:**
- [Hetzner Cloud](https://www.hetzner.com/cloud)
- [Hetzner Referral Program Terms](https://www.hetzner.com/legal/referrals)
- [Hetzner Community Tutorials](https://community.hetzner.com/tutorials/)
- [Hetzner Community Content (GitHub)](https://github.com/hetzneronline/community-content)
- [LRob: EUR 20 Free Hetzner Cloud](https://www.lrob.fr/en/blog/20-euros-free-hetzner-cloud/)

## Submission 96: Google Colab Pro Free for US Students & Faculty

- Category: Education & Student
- Source Link: https://blog.google/products-and-platforms/products/education/colab-higher-education
- Amount: 1 year of Colab Pro (normally $11.99/month, ~$144/year value)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Google offers **free 1-year Colab Pro subscriptions** to students and faculty at US-based higher education institutions. This is the full Colab Pro plan (normally $11.99/month, ~$144/year value) with premium GPU access (T4, L4, A100), extended session runtimes, increased memory, and a monthly allowance of compute units. Verified through SheerID. Available while supplies last.

---

## Registration (Step-by-Step)

1. Go to the [Colab Pro signup page](https://colab.research.google.com/signup)
2. Sign in with your **Google account** (the same one you use for Colab)
3. Look for the **"No cost to students and educators"** button
4. Click it to begin the **SheerID verification** process
5. Fill out the SheerID form with:
   - Your **name** and **institutional email**
   - Select your **US-based university** from the dropdown
6. **Critical:** Use the **same email address** as your Colab/Google account in the SheerID form — if you enter your .edu email but are logged into a different Google account, verification will not be recorded
7. SheerID will either **instantly verify** you or request additional documentation
8. Once verified, your free Colab Pro subscription activates immediately

**Caveats:**
- You must be **located in the US** at the time of verification (VPNs may cause issues)
- Available **while supplies last** — Google has not disclosed how many subscriptions are available
- SheerID will communicate via email until verification is complete

---

## What You Get

You get the **exact same benefits** as a paid Colab Pro subscription:

| Feature | Details |
|---|---|
| **Premium GPUs** | Priority access to T4 GPU, potential access to L4 and A100 GPUs |
| **Extended Runtimes** | Longer notebook session times (up to ~24 hours vs ~12 for free tier) |
| **More Memory** | High-memory VM option for larger datasets and models |
| **Compute Units** | Monthly allowance of compute units (100 CU/month) |
| **Background Execution** | Continue running notebooks with the browser closed |
| **Terminal Access** | Built-in terminal in notebooks |

### GPU Performance Reference

| GPU | Approx. CU/hour | Best For |
|---|---|---|
| **T4** | ~11.7 CU/hr | Standard ML tasks, inference, fine-tuning small models |
| **L4** | ~20 CU/hr | Complex neural networks, large image processing |
| **A100** | ~62 CU/hr | Large-scale deep learning, training large models |

### Additional Education Features

- **Slideshow Mode:** Turn notebooks into interactive presentations with live code execution — great for teaching
- **AI Feature Controls:** Instructors can toggle AI features on/off per notebook to ensure students learn fundamentals

---

## Eligibility & Verification

- **Who qualifies:** Students AND faculty at **US-based** higher education institutions
- **Location requirement:** Must be physically located in the US at the time of verification
- **Verification method:** SheerID (third-party identity verification)
- **Account requirement:** Must have a Google account
- **Supply-limited:** Available while spots remain — not guaranteed indefinitely

---

## Renewal

- The free subscription lasts **1 year** from activation
- You can **re-verify** your student or faculty status **335 days** after initial signup
- Re-verification extends your subscription for another year if you are still eligible
- No auto-renewal to paid plan — when your free year ends, you simply revert to the free Colab tier

---

## Additional Tips

- **Compute units are shared** — if you burn through your monthly CU allowance on A100 sessions, you will be throttled back to free-tier resources for the rest of the month
- **GPU availability is not guaranteed** — even with Pro, you are not guaranteed an A100. GPUs are allocated based on demand and your remaining CU balance
- **Save your work frequently** — sessions can still disconnect due to inactivity or resource limits
- **Pair with Google Drive** for persistent storage — Colab VMs are ephemeral
- **Related offer:** Google AI Pro (Gemini Advanced) is also free for students globally — separate program, can be used alongside Colab Pro
- **Alternative GPU options:** If you need more guaranteed GPU time, also check Kaggle (30 hrs/week free), Lightning AI, SageMaker Studio Lab, or Paperspace Gradient
- This program was announced in **July 2025** by Google

---

**Sources:**
- [Google Blog — Colab for Higher Education](https://blog.google/products-and-platforms/products/education/colab-higher-education/)
- [Colab Pro Signup Page](https://colab.research.google.com/signup)
- [Google Colab FAQ](https://research.google.com/colaboratory/faq.html)
- [Jeff Dean on X — Colab Pro Announcement](https://x.com/JeffDean/status/1946437195121963277)
- [Talk Android — Colab Pro Free Education](https://www.talkandroid.com/512277-google-collab-pro-free-education/)

## Submission 95: Google AI Pro Free for Students (1 Year)

- Category: Education & Student
- Source Link: https://gemini.google/students
- Amount: 1 year of Google AI Pro (valued ~$240/year)
- Difficulty: 2/5 (Easy)
- Expires At: 2026-04-30

### Description

Google offers eligible university students **12 months of Google AI Pro (formerly Google One AI Premium)** completely free. This gives you full access to **Gemini Advanced** with Gemini 2.5 Pro, Deep Research, Veo 2 video generation, NotebookLM Plus, Gemini in Google Workspace (Gmail, Docs, Sheets, Slides), and 2TB of Google One cloud storage. The plan is normally $19.99/month (~$240/year). Available in **120+ countries**, verified through SheerID.

---

## Registration (Step-by-Step)

1. Go to [gemini.google/students](https://gemini.google/students)
2. **Sign in with a personal Gmail account** — school Google Workspace accounts do NOT qualify
3. Click **"Start trial"** or equivalent signup button
4. You will be redirected to **SheerID** for student verification
5. Select your institution from the dropdown and enter your **.edu email** (or equivalent academic email)
6. **Instant verification path:** If your school email is in SheerID's database, you receive a verification link within minutes
7. **Document upload path:** If instant verification fails, upload one of the following (dated within 90 days):
   - Current class schedule
   - Tuition receipt
   - Enrollment letter
   - Student ID card with current academic year
8. **Add a payment method** (credit card, debit card, or PayPal) — you will NOT be charged during the trial
9. Complete the subscription flow through Google Play Store
10. Done — your 12-month free trial begins immediately

**Critical caveats:**
- Your **name must match** between your Google account and SheerID verification exactly
- **Disable VPNs** and clear browser cache if you get location errors
- If your school shows as "not found," try alternate spellings or abbreviations
- SheerID manual review (document upload) can take **up to 48 hours**
- SheerID support typically responds within 24 business hours if you have issues

---

## What You Get

| Feature | Details |
|---|---|
| **Gemini Advanced** | Full access to Gemini 2.5 Pro with 1.5 million token context window |
| **Deep Research** | AI-powered literature reviews and multi-step research |
| **Veo 2** | AI video generation |
| **NotebookLM Plus** | Enhanced AI notebook with audio overviews and mind maps |
| **Gemini in Workspace** | AI assistance in Gmail, Docs, Sheets, and Slides |
| **Google One Storage** | 2TB cloud storage (shared across Drive, Gmail, Photos) |
| **Image Generation** | Pro-level Imagen image generation |
| **Custom Gems** | Create personalized AI assistants |

---

## Eligibility & Verification

- **Age:** Must be 18 or older
- **Enrollment:** Actively enrolled at an accredited higher education institution (university or college)
- **Part-time students:** Qualify
- **International students:** Qualify based on enrollment status, not citizenship (must be enrolled at a school in an eligible country)
- **High school students:** Do NOT qualify
- **Recent graduates/alumni:** Do NOT qualify
- **Verification method:** SheerID (instant via .edu email or manual document upload)

### Supported Countries (120+)

**North America:** USA, Canada | **Europe:** UK, Germany, France, Spain, Italy, Netherlands, and most EU nations | **Asia:** Japan, South Korea, India, Indonesia, Malaysia, Singapore, Philippines, Thailand, Vietnam | **Middle East/Africa:** Egypt, Saudi Arabia, UAE, South Africa, Nigeria, Kenya, Ghana, Rwanda, Zimbabwe

**Not supported:** Mainland China, Russia, and countries with restricted Google services

### Sign-up Deadline

The current deadline is **April 30, 2026**. Check the signup page for the latest deadline as Google has extended it multiple times.

---

## Auto-Renewal & Cancellation

- After 12 months, the plan **auto-renews at $19.99/month**
- Cancel at least **1 day before** trial ends at [one.google.com/settings](https://one.google.com/settings)
- You keep access until the trial period officially ends, even after canceling
- After cancellation: Gemini reverts to free tier, NotebookLM Plus reverts to basic, storage above 15GB requires a paid Google One plan
- **Set a calendar reminder 1-2 weeks before your trial expires** to avoid surprise charges

---

## Additional Tips

- **Personal Gmail only** — this is the #1 mistake. School Google Workspace accounts cannot be used, even if they end in a recognized domain
- Students in select countries (US, UK, Japan, Indonesia, Brazil) who activated early may be able to **re-verify for an additional 12 months**
- **Do not share your account** — Google monitors for multi-user access and can terminate the subscription
- Deep Research is great for literature reviews but always **verify sources independently** — it can hallucinate citations
- After the trial ends, free alternatives include: Gemini Free, ChatGPT Free, Claude Free, Microsoft Copilot
- **Related program:** Google Colab Pro is also free for US students and faculty (separate offer)

---

**Sources:**
- [Google AI Pro Student Trial — Official Page](https://gemini.google/students/)
- [Google Support — Student Trial Details](https://support.google.com/gemini/answer/16417758?hl=en)
- [AI Free API — Gemini Student Plan Guide](https://www.aifreeapi.com/en/posts/gemini-student-plan)
- [Tom's Guide — Gemini AI Pro Free for Students](https://www.tomsguide.com/ai/gemini-ai-pro-is-free-for-students-for-one-year-heres-how-to-get-the-gemini-3-upgrade-right-now)
- [Student Discount IO — Google AI Pro Student](https://studentdiscount.io/blog/google-ai-pro-student-free)

## Submission 94: Gemini API Free Tier (Google AI Studio)

- Category: AI API Free Tiers
- Source Link: https://ai.google.dev/gemini-api/docs/pricing
- Amount: Free ongoing: 5-15 RPM, 100-1,000 RPD depending on model. 1M token context window. All Gemini models included.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Google offers **completely free access** to its Gemini model family via API -- no credit card required. This includes Gemini 2.5 Pro (their most capable model), Gemini 2.5 Flash, Flash-Lite, and open models like Gemma 3. All models support a **1M token context window** on the free tier. The catch: rate limits are tight (especially after the December 2025 cuts), and Google may use your free-tier data for model training.

## Registration (Step-by-Step)

1. Go to [Google AI Studio](https://aistudio.google.com/) and sign in with your Google account.
2. Click **Get API Key** to generate a free API key -- no billing setup needed.
3. Use the key with the Gemini API (REST or official SDKs for Python, Node.js, Go, Dart, etc.).
4. You can test prompts directly in AI Studio's web playground before writing code.
5. Rate limits apply per **Google Cloud project**, not per API key.

## Available Models (Free Tier)

| Model | Best For |
|-------|----------|
| Gemini 2.5 Pro | Most capable -- complex reasoning, coding, analysis |
| Gemini 2.5 Flash | Fast + smart -- balanced speed/quality |
| Gemini 2.5 Flash-Lite | Highest free-tier limits -- lightweight tasks |
| Gemini 2.0 Flash | Previous-gen fast model |
| Gemini 3 Flash Preview | Latest preview model |
| Gemini 3.1 Pro Preview | Frontier preview (limited availability) |
| Gemma 3 / Gemma 3n | Open models, fully free |
| Gemini Embedding | Text embeddings |

## Free Tier Rate Limits

| Model | RPM | RPD | TPM |
|-------|-----|-----|-----|
| Gemini 2.5 Pro | 5 | 100 | 250,000 |
| Gemini 2.5 Flash | 10 | 250 | 250,000 |
| Gemini 2.5 Flash-Lite | 15 | 1,000 | 250,000 |
| Gemini 2.0 Flash | 5 | -- | 250,000 |
| Gemini 3 Flash Preview | Free | Free | Free |
| Gemma 3 / Gemma 3n | Free | Free | Free |

*RPM = Requests per Minute, RPD = Requests per Day, TPM = Tokens per Minute*

## Additional Features (Free)

| Feature | Availability |
|---------|-------------|
| Context window | 1M tokens (all models) |
| Google Search grounding | 500-5,000 RPD depending on model |
| Google Maps grounding | 500-10,000 RPD depending on model |
| Context caching | Available on select models |
| Multimodal input | Images, audio, video, PDF |
| Structured output (JSON) | Supported |
| Function calling | Supported |

## Additional Tips

- **December 2025 quota cuts:** Google reduced free-tier limits by 50-80% in December 2025. Flash went from ~250 RPD to as low as 50 in some configs. Optimize your usage.
- **Data usage warning:** On the free tier, Google states that your input and output data may be used to improve their products. Switch to the paid tier to opt out.
- **Daily quotas reset at midnight Pacific Time.** Plan batch jobs accordingly.
- **Limits are per project, not per key.** Creating multiple API keys in the same project does not increase your quota.
- **Flash-Lite is your best friend** on the free tier -- 15 RPM and 1,000 RPD is generous enough for most prototyping and small-scale apps.
- **Gemini 2.5 Pro at 5 RPM / 100 RPD** is tight but usable for development. Great for complex one-off tasks, not for serving traffic.
- **Multimodal is free** -- you can process images, audio, and video with no additional cost on the free tier.
- **Pair with Vertex AI** for higher limits: Google Cloud's Vertex AI offers Gemini with higher rate limits under the $300 Google Cloud free trial.

**Sources:**
- https://ai.google.dev/gemini-api/docs/pricing
- https://ai.google.dev/gemini-api/docs/rate-limits
- https://aistudio.google.com/
- https://www.aifreeapi.com/en/posts/gemini-api-free-tier-rate-limits

## Submission 93: Fluidstack Startup GPU Credits - Up to $100,000

- Category: Free GPU & Compute
- Source Link: https://www.fluidstack.io/nvidia_inception_offering
- Amount: Up to $100,000 in GPU compute credits (covers compute, storage, and network)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Fluidstack offers **up to $100,000 in compute credits** for NVIDIA Inception member startups, covering H100, H200, and A100 GPU usage. Credits also cover storage and network costs, so you can scale from prototype to production without surprise charges. Fluidstack aggregates underutilized GPUs from data centers worldwide, offering competitive pricing (H100 from $2.10/hr, A100 from $1.30/hr). The program includes a dedicated technical account manager. Fluidstack is growing fast — reportedly in talks with Google for a $100M investment and collaborating with Anthropic on a $50B data center initiative.

---

## Registration (Step-by-Step)

1. **Join NVIDIA Inception** — apply at [nvidia.com/en-us/startups](https://www.nvidia.com/en-us/startups/). The program is free, open to all stages, with no application fees or deadlines
2. Your company must be incorporated, employ at least one developer, have a working website, and be less than 10 years old
3. Once accepted into Inception, go to [fluidstack.io/nvidia_inception_offering](https://www.fluidstack.io/nvidia_inception_offering) and apply for the Fluidstack credits
4. Provide details about your AI product, compute needs, and how you plan to use the credits
5. Fluidstack reviews your application — credit amounts are based on your adoption of NVIDIA technologies and intended use
6. Once approved, credits are applied to your Fluidstack account covering compute, storage, and network
7. A dedicated technical account manager is assigned to help you get started

**Caveats:**
- Credit allocation is not guaranteed at the full $100,000 — amount depends on your company profile and NVIDIA technology adoption
- You must be an active NVIDIA Inception member
- Credits cover compute, storage, and network costs (no hidden fees)
- No long-term commitments required

---

## GPU Hardware & Specs

Fluidstack offers dedicated GPU instances optimized for AI workloads:

| GPU | VRAM | Memory Type | Price |
|---|---|---|---|
| **A100 PCIe** | 40 GB | HBM2e | $1.80/hr |
| **A100 SXM** | 80 GB | HBM2e | $1.30/hr |
| **L40S** | 48 GB | GDDR6X | $1.30/hr |
| **H100** | 80 GB | HBM3 | $2.10/hr |
| **H200** | 141 GB | HBM3e | $2.30/hr |

*All instances are dedicated (not shared). Private clusters available on request for larger deployments.*

---

## What $100,000 Gets You (Rough Estimates)

| GPU | Approx. Hours |
|---|---|
| A100 SXM (80 GB) | ~76,900 hours |
| L40S (48 GB) | ~76,900 hours |
| H100 (80 GB) | ~47,600 hours |
| H200 (141 GB) | ~43,500 hours |

---

## Free Tier Limits

| Benefit | Detail |
|---|---|
| Credit amount | Up to $100,000 |
| GPU access | H100, H200, A100, L40S |
| Covers | Compute + storage + network |
| Technical support | Dedicated account manager |
| Contracts | No long-term commitment |
| Infrastructure | Managed Kubernetes and SLURM environments |

---

## Additional Tips

- **NVIDIA Inception is the key** — you must be an Inception member first. Inception is free and also unlocks credits from dozens of other cloud providers (Nebius, AWS, Azure, GCP, etc.)
- **Credit amount depends on your profile** — cloud credit requests are evaluated based on your company's adoption of NVIDIA technologies, intended use for solution development, and completeness of your Inception portal profile. Fill out your profile thoroughly
- **Fluidstack is price-competitive** — H100 at $2.10/hr and A100 SXM at $1.30/hr are among the lowest rates in the market, comparable to Vast.ai and cheaper than major clouds
- **Dedicated account manager is valuable** — unlike self-serve platforms, you get a real human to help optimize your setup. Use this for cluster configuration and cost optimization advice
- **Stack with other Inception benefits** — NVIDIA Inception gives you credits across multiple providers. You can use Fluidstack for training and a different provider for inference
- **Managed infrastructure available** — Fluidstack offers managed Kubernetes and SLURM environments, which saves significant DevOps time for distributed training
- **Company is well-funded and growing** — Fluidstack is reportedly raising at a $7B+ valuation with Google backing, so the platform is likely to be stable and well-supported

---

**Sources:**
- [Fluidstack x NVIDIA Inception Offering](https://www.fluidstack.io/nvidia_inception_offering)
- [Fluidstack GPU Pricing — ComputePrices.com](https://computeprices.com/providers/fluidstack)
- [Fluidstack Review — GetDeploying](https://getdeploying.com/fluidstack)
- [NVIDIA Inception Program](https://www.nvidia.com/en-us/startups/)
- [NVIDIA Inception Guide — ThunderCompute](https://www.thundercompute.com/blog/nvidia-inception-program-guide)
- [Google $100M Fluidstack Investment — SiliconANGLE](https://siliconangle.com/2026/02/20/google-reportedly-invest-100m-ai-cloud-operator-fluidstack/)

## Submission 92: fal.ai Free Credits for Image/Video Generation

- Category: AI API Free Tiers
- Source Link: https://fal.ai/pricing
- Amount: Free credits on signup (amount varies). 600+ models. Pay-per-use from $0.01/image.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

fal.ai is a **generative AI API platform** offering **600+ models** for image, video, 3D, and audio generation with a **pay-per-use** pricing model. New users get **free credits** to explore the platform (exact amount varies by promotional period). The platform is developer-focused with fast inference (claims 4x faster than competitors), an API-first approach, and models like **FLUX**, **Kling**, **Veo**, **Hailuo**, and **Seedream**. Standard users get **2 concurrent tasks**, while users with $1,000+ in credits get **40 concurrent tasks**. Pricing is transparent and output-based (per image/video), not per-token.

---

## Registration (Step-by-Step)

1. Go to [fal.ai](https://fal.ai) and click **Sign Up**
2. Create an account using email or GitHub
3. You receive free credits upon signup (amount varies by current promotion)
4. Navigate to the **Dashboard** to view your credit balance
5. Generate an **API key** from your account settings
6. Install the fal client: `pip install fal-client` (Python) or `npm install @fal-ai/client` (JavaScript)
7. Start making API requests to any of the 600+ models

**Important:**
- Free credits expire based on the type of grant (varies from 1 week to never)
- Purchased credits expire in 365 days from purchase
- Failed requests due to user error (HTTP 422) are still charged; server errors (5xx) are not
- Cold start is very rare on main endpoints — you won't be charged for startup time

---

## Available AI Models / Services

### Image Generation

| Model | Price/Image (1MP) | Notes |
|---|---|---|
| **Seedream V4** | $0.03 | 33 images per $1 |
| **FLUX Kontext Pro** | $0.04 | 25 images per $1 |
| **Nanobanana** | $0.04 | 25 images per $1 |
| **Qwen** | $0.02/MP | 50 megapixels per $1 |
| **FLUX 1.1** | $0.04 | Standard text-to-image |
| **HiDream** | $0.01 | Budget image generation |
| **Venice SD35** | $0.01 | Budget Stable Diffusion |
| **Recraft V4 Pro** | $0.29 | Premium quality |

### Video Generation

| Model | Price/Second | Notes |
|---|---|---|
| **Wan 2.5** | $0.05/sec | 20 seconds per $1 |
| **Kling 2.5 Turbo Pro** | $0.07/sec | 14 seconds per $1 |
| **Ovi** | $0.20/video | 5 videos per $1 |
| **Veo 3** | $0.40/sec | 3 seconds per $1 (premium) |
| **Google Veo 2** | $2.50/5sec | Free preview available |

### GPU Compute (for custom models)

| GPU | VRAM | Price/Hour |
|---|---|---|
| **A100** | 40GB | $0.99/h |
| **H100** | 80GB | $1.89/h |
| **H200** | 141GB | $2.10/h |
| **B200** | 184GB | Contact sales |

---

## Free Tier Limits

| Feature | Standard | Premium ($1,000+ credits) |
|---|---|---|
| Free credits on signup | Yes (amount varies) | — |
| Concurrent tasks | 2 | 40 |
| Queue behavior | Excess requests queued | Higher priority |
| API priority | Below playground | Standard |
| Coupon credits | Playground/Sandbox only | Full API access |
| Credit expiration | Varies (1 week to never) | 365 days from purchase |

---

## Additional Tips

- **Output-based pricing** — most models charge per image or per video second, not per token or GPU hour. This makes costs predictable and easy to estimate
- **FLUX 1.1 at $0.04/image** is one of the best value propositions for high-quality text-to-image generation
- **Free coupon credits** are redeemable at `fal.ai/coupon-claim/COUPON-CODE` but can only be used in the **Sandbox and Playground**, not via API or workflows
- **Watch for promotions** — fal.ai regularly gives away free credits on social media (Twitter/X) to celebrate new model launches. Follow [@FAL](https://x.com/FAL) for announcements
- **Invoice-based payments** available for higher-volume customers upon request
- **Volume discounts** available upon request for teams with significant usage
- **Resolution affects price** — image prices scale proportionally with resolution. A 2MP image costs roughly 2x a 1MP image
- **Video length affects price** — video costs scale with duration. Keep generations short during experimentation
- **API-first design** — fal.ai is built for developers. The playground is useful for testing, but the real value is in the API
- **Compare with alternatives** — for image generation APIs, also check Replicate, Leonardo AI, and Stability AI. For video, compare with RunwayML and Pika

---

**Sources:**
- [fal.ai](https://fal.ai)
- [fal.ai Pricing](https://fal.ai/pricing)
- [fal.ai Documentation - Pricing](https://docs.fal.ai/platform-apis/v1/models/pricing)
- [fal.ai FAQ](https://docs.fal.ai/model-apis/faq)
- [fal.ai on Twitter/X](https://x.com/FAL)

## Submission 91: ElevenLabs Free Text-to-Speech API

- Category: AI API Free Tiers
- Source Link: https://elevenlabs.io/pricing/api
- Amount: 10,000 characters/month (~10 min audio). 2,500 char/request limit. No commercial use.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

ElevenLabs offers one of the best text-to-speech APIs on the market, and their free tier gives you 10,000 characters per month (~10 minutes of audio). No credit card required. Supports dozens of languages with natural-sounding voices. The catch: no commercial use and no voice cloning on the free plan.

## Registration (Step-by-Step)

1. Go to [elevenlabs.io](https://elevenlabs.io) and click **Sign Up**.
2. Create an account with email, Google, or Apple.
3. Verify your email address.
4. Navigate to **Profile > API Keys** or go to [elevenlabs.io/app/settings/api-keys](https://elevenlabs.io/app/settings/api-keys) and generate your API key.
5. Use the API at `https://api.elevenlabs.io/v1/text-to-speech/{voice_id}` with your key in the `xi-api-key` header.
6. Explore available voices in the **Voice Library** within the dashboard.

**Caveat:** The voice library is not accessible via API on the free tier. You'll need to browse voices in the web UI and note their IDs.

## Available AI Models / Services

| Feature | Free Tier Access |
|---|---|
| Text-to-Speech (TTS) | Yes -- 10,000 chars/month |
| Multilingual TTS | Yes -- dozens of languages |
| Voice Library (web) | Yes -- browse and use pre-made voices |
| Voice Library (API) | No -- paid plans only |
| Custom Voice Cloning | No -- paid plans only |
| Professional Voice Cloning | No -- paid plans only |
| Commercial Usage Rights | No -- paid plans only |
| Projects (long-form audio) | Limited |
| Sound Effects Generation | Limited |

Supported languages include English, Spanish, French, German, Mandarin, Japanese, Korean, Portuguese, Italian, and many more.

## Free Tier Limits

| Resource | Limit |
|---|---|
| Characters per month | 10,000 |
| Characters per request | 2,500 |
| Audio output | ~10 minutes/month |
| Custom voices | Up to 3 |
| Voice cloning | Not available |
| Commercial use | Not permitted |
| API access | Available (with limitations) |

## Additional Tips

- **10,000 chars goes fast:** That's roughly 1,500-2,000 words or ~10 minutes of audio. Plan your usage if you're prototyping.
- **2,500 char per-request limit:** Split longer texts into chunks. Most TTS wrappers handle this automatically.
- **No commercial use on free tier:** You can experiment and build prototypes, but you can't use the generated audio in any commercial product, video, or app without upgrading.
- **Voice IDs:** Browse voices at [elevenlabs.io/voice-library](https://elevenlabs.io/voice-library), click a voice, and grab the ID from the URL or API.
- **Starter plan is $5/month:** If you outgrow the free tier, the cheapest paid plan gives you 30,000 chars/month with commercial rights for just $5.
- **WebSocket streaming:** ElevenLabs supports real-time TTS streaming via WebSocket for low-latency applications.
- **SDKs available:** Official Python and JavaScript/TypeScript SDKs. Community SDKs for other languages.
- **Unified character credits:** As of 2025, ElevenLabs unified billing so all models (Multilingual, Flash) cost the same number of character credits.
- **Quality is top-tier:** ElevenLabs consistently ranks among the most natural-sounding TTS APIs. The free tier uses the same models as paid plans.

**Sources:**
- https://elevenlabs.io/pricing/api
- https://elevenlabs.io/pricing
- https://elevenlabs.io/docs/overview/capabilities/text-to-speech
- https://voice.ai/hub/tts/elevenlabs-text-to-speech/

## Submission 90: ElevenLabs AI Student Pack ($1,500+ in Credits)

- Category: Education & Student
- Source Link: https://aistudentpack.com/
- Amount: $1,500+ combined value across 16 tools (mix of free subscriptions, credits, and discounts)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

The **AI Student Pack** by ElevenLabs bundles **$1,500+ in free credits and discounts across 16 AI tools** — covering voice synthesis, note-taking, presentations, AI coding, video generation, and more. It is one of the highest-value student bundles in the AI space. **Volume 1 is fully claimed.** Volume 2 is coming — you can join the waitlist at [aistudentpack.com](https://aistudentpack.com). Announced August 2025.

---

## Registration (Step-by-Step)

**Volume 1 (CLOSED):**
1. Go to [aistudentpack.com](https://aistudentpack.com)
2. Log in with a **verified student email** (.edu or equivalent academic email)
3. Browse the 16 partner offers
4. Claim each offer individually — methods vary:
   - **Fetch/auto-claim:** Click to reveal your code or activate directly
   - **Promo code:** Copy and apply at the partner's checkout
   - **External link:** Redirects to partner site for activation
5. Some offers require creating an account with the partner service first

**Volume 2 (UPCOMING):**
1. Go to [aistudentpack.com](https://aistudentpack.com)
2. Click **"Join Waitlist"**
3. Enter your email to be notified when Volume 2 launches
4. No confirmed launch date yet

---

## What You Get (Volume 1 Offers)

### Free Subscriptions

| Tool | Offer | Value | Category |
|---|---|---|---|
| **ElevenLabs** | 3 months free Creator Plan | ~$55 | Voice AI / TTS |
| **ElevenReader** | 1 year free Plus | ~$96 | Text-to-speech reader |
| **Granola** | 6 months free Individual | ~$85 | AI meeting notes |
| **Notion** | Notion Plus free | ~$120 | Productivity / notes |

### Credits

| Tool | Offer | Value | Category |
|---|---|---|---|
| **V0** (Vercel) | $60 in credits | $60 | AI UI generation |

### Discounts (50%+)

| Tool | Offer | Value | Category |
|---|---|---|---|
| **Headspace** | 85% off | ~$65 saved | Meditation / wellness |
| **Superhuman** | 67% off Starter | ~$240 saved | AI email client |
| **Bolt** | 60% off Pro for 1 year | ~$140 saved | AI app builder |
| **Blinkist** | 60% off yearly | ~$100 saved | Book summaries |
| **Coconote** | 50% off first year | ~$65 saved | AI notes |
| **Hedra** | 50% off plans | Up to $450 saved | AI video avatars |
| **HeyGen** | 50% off Creator | ~$170 saved | AI video generation |
| **Gamma** | 50% off plans | ~$90 saved | AI presentations |
| **Lovable** | 50% off Pro for 1 year | ~$150 saved | AI app builder |
| **Sketchy** | 1 month free | ~$20 | Medical study tool |
| **Higgsfield** | 20% off first month | ~$30 saved | AI video |

### Known Promo Codes (Volume 1)

- **Blinkist:** `keeplearning_elevenlabspromo60`
- **Coconote:** `ELEVEN50`
- **Hedra:** `AISTUDENTPRO` or `AISTUDENTCREATOR`
- **Gamma:** `MBPT4UY`
- **Sketchy:** `ELEVENLABS`
- **Higgsfield:** `STUDENT20`

*Note: These codes were for Volume 1 and may not work for Volume 2.*

---

## Eligibility & Verification

- **Students** with a verified student email (.edu or equivalent academic email)
- Verification appears to be **email-based** — no SheerID or document upload required
- No explicit age restrictions mentioned
- No geographic restrictions mentioned — appears to be **globally available**
- Volume 1 was **supply-limited** (all codes were claimed). Volume 2 is expected to have similar limits

---

## Additional Tips

- **Claim fast when Volume 2 drops** — Volume 1 sold out quickly. Join the waitlist and act immediately when notified
- **The total "$1,500+" value is a mix of free and discounted offers** — not all are fully free. About $400-500 is in genuine free subscriptions/credits; the rest is in percentage discounts
- **Each tool requires separate activation** — there is no single bundle redemption
- **Some offers require a credit card** for the discounted subscription and will auto-renew at full price — set cancellation reminders
- **V0 credits ($60) and ElevenLabs Creator Plan (3 months) are the most universally useful** for builders and indie devs
- **Bolt and Lovable** (both AI app builders) are great for vibe coders who want to prototype quickly
- **Related packs:** ElevenLabs also runs an [AI Engineer Pack](https://www.aiengineerpack.com/) and an [AI Creator Pack](https://www.aicreatorpack.com/) with different tool selections
- **Compare with GitHub Student Developer Pack** — the GitHub pack offers more developer tools while the AI Student Pack focuses on AI-specific creative and productivity tools

---

**Sources:**
- [AI Student Pack — Official Site](https://aistudentpack.com/)
- [ElevenLabs Blog — AI Student Pack Announcement](https://elevenlabs.io/blog/ai-student-pack)
- [ElevenLabs for Students](https://elevenlabs.io/students)
- [ElevenLabs on X — AI Student Pack Launch](https://x.com/elevenlabsio/status/1956023074546921587)
- [Cloud Credits IO — AI Student Pack Details](https://cloudcredits.io/providers/elevenlabs/programs/ai-student-pack)

## Submission 89: Deepgram $200 Free Speech API Credits

- Category: AI API Free Tiers
- Source Link: https://deepgram.com/pricing
- Amount: $200 free credits (never expire). ~25,000+ min transcription at PAYG rates.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Deepgram gives every new user $200 in free credits for their speech-to-text, text-to-speech, voice agent, and audio intelligence APIs. No credit card required. Credits never expire. At pay-as-you-go rates, that's roughly 25,000+ minutes of transcription or 6.6M+ characters of text-to-speech.

## Registration (Step-by-Step)

1. Go to [console.deepgram.com](https://console.deepgram.com) and click **Sign Up**.
2. Create an account with email, Google, or GitHub.
3. You'll immediately receive **$200 in free credits** -- no credit card needed.
4. Navigate to **API Keys** in your dashboard and create a key.
5. Start making API calls to `https://api.deepgram.com/v1/` with your key.
6. Credits never expire, so take your time experimenting.

**Caveat:** If you later add a payment method with auto-reload enabled, the default is $100 reload when balance drops below $10. Disable auto-reload if you only want to use free credits.

## Available AI Models / Services

### Speech-to-Text (STT)
| Model | Best For | Price/min (PAYG) |
|---|---|---|
| Flux | Fastest, real-time streaming | $0.0077 |
| Nova-3 Monolingual | Best English accuracy | $0.0077 |
| Nova-3 Multilingual | 40+ languages | $0.0092 |
| Nova-2 | Legacy, still solid | $0.0058 |
| Nova-1 | Budget option | $0.0058 |
| Enhanced | Older, higher quality | $0.0165 |
| Base | Cheapest legacy model | $0.0145 |

### Text-to-Speech (TTS)
| Model | Price/1K chars (PAYG) |
|---|---|
| Aura-2 | $0.030 |
| Aura-1 | $0.015 |

### Voice Agent API
| Tier | Price/min |
|---|---|
| Full-stack (Deepgram LLM + TTS) | $0.08 - $0.16 |
| Bring-your-own LLM + TTS | $0.04 |

### Add-on Features
| Feature | Price/min |
|---|---|
| Speaker Diarization | $0.0020 |
| Redaction (PII) | $0.0020 |
| Keyterm Prompting | $0.0013 |
| Smart Formatting | Included |

## Free Tier Limits

| Resource | Limit |
|---|---|
| Free credits | $200 |
| Credit expiry | Never |
| STT concurrency (REST) | 100 |
| STT concurrency (WebSocket) | 150 |
| TTS concurrency | 45 |
| Voice Agent concurrency | 45 |
| Whisper Cloud concurrency | 5 |
| Billing granularity | Per-second (no rounding) |

## Additional Tips

- **Credits go far:** At Nova-3 rates ($0.0077/min), $200 = ~25,974 minutes (~433 hours) of transcription. With Nova-2 ($0.0058/min), it's ~34,483 minutes (~575 hours).
- **True per-second billing:** Unlike competitors that round to the nearest minute, Deepgram bills per-second. A 10-second clip costs what it should.
- **Multichannel pricing:** Each audio channel is billed separately. Stereo = 2x the single-channel rate.
- **WebSocket streaming:** Real-time transcription via WebSocket is the same price as REST -- no streaming surcharge.
- **Auto-reload gotcha:** When you add a payment method, auto-reload defaults to ON ($100 when balance < $10). Turn it off in settings if you want to stay on free credits only.
- **Refund policy:** Unused purchased credits (not free credits) can be refunded within 30 days.
- **SDKs available:** Official SDKs for Python, Node.js, Go, .NET, and Rust.

**Sources:**
- https://deepgram.com/pricing
- https://deepgram.com/learn/speech-to-text-api-pricing-breakdown-2025
- https://brasstranscripts.com/blog/deepgram-pricing-per-minute-2025-real-time-vs-batch

## Submission 88: Databricks Free Edition & University Alliance

- Category: Education & Student
- Source Link: https://www.databricks.com/university
- Amount: Free access to full Databricks platform (Free Edition) + free courses + 25% off certifications
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Databricks offers its full Data Intelligence Platform free to everyone (students, hobbyists, professionals) through Databricks Free Edition -- no cloud account or credit card needed. For students at 1,200+ partner universities, the University Alliance program adds structured learning paths, certification discounts, and hands-on experience with industry tools. Part of Databricks' $100M investment in AI/data education.

## Registration (Step-by-Step)

### Free Edition (Anyone)
1. Go to [databricks.com/university](https://www.databricks.com/university) or directly to the Free Edition signup.
2. Create an account -- **no credit card required**.
3. You'll get a workspace with serverless compute, ready to use immediately.
4. Start with notebooks (Python or SQL), explore dashboards, or build AI agents.

### University Alliance (Students at Partner Schools)
1. Check if your school is part of the University Alliance (1,200+ institutions including Duke, Georgia Tech, UC Berkeley, Columbia, Texas A&M, and many more).
2. Go to [databricks.com/resources/student-resources](https://www.databricks.com/resources/student-resources).
3. Sign up with your student email.
4. Access course-specific workspaces your professors may have set up.
5. Take free self-paced courses on Databricks Academy.
6. After completing courses, get **25% off Databricks certifications**.

## What You Get

### Free Edition Features
- **Collaborative Notebooks** -- Python and SQL support with real-time collaboration.
- **SQL Editor** -- query and analyze data directly.
- **AI/ML Tools** -- Mosaic AI for building AI agents and applications, experiment with foundation models.
- **Dashboards & Genie** -- AI-powered business intelligence with natural language queries.
- **Data Engineering** -- Lakeflow for building data pipelines (ingestion, transformation, orchestration).
- **Databricks Assistant** -- in-product AI coding help that suggests, explains, and fixes code.
- **Apache Spark, Delta Lake, MLflow** -- industry-standard tools for data processing and ML.

### Free Edition Limitations
- **Serverless compute only** -- no provisioned clusters or GPUs.
- **Daily usage limits** -- compute and storage are capped daily; pauses until reset if exceeded.
- **No Scala or RDDs** -- Python and SQL only.
- **No commercial use** -- Free Edition accounts are for learning and personal projects.
- **Small cluster sizes** -- one SQL warehouse limited to 2X-Small.
- **Limited outbound internet** -- restricted network access from notebooks.
- **Inactive accounts may be deactivated** after prolonged inactivity.

### University Alliance Extras
- **Free self-paced courses** on Databricks Academy (hundreds of hours).
- **Role-based learning paths:** Data Engineering, Data Analysis, Machine Learning, Generative AI.
- **25% off certifications** after completing a Databricks course.
- **Curriculum resources** for educators (pre-built labs, assignments).

## Eligibility & Verification

- **Free Edition:** Open to everyone globally. No verification needed beyond email signup.
- **University Alliance:** Students and educators at 1,200+ partner institutions worldwide.
- **Certification discount:** Requires completion of at least one Databricks course.

## Additional Tips

- **Community Edition is being discontinued** -- if you were using it, migrate to Free Edition now.
- **No credit card, no expiration** -- Free Edition is genuinely free with no time limit (unlike trial programs).
- **Great for learning Spark** -- if you're studying for data engineering roles, hands-on Spark/Delta Lake experience is highly valued by employers.
- **Pair with cloud credits** -- if you need more compute than Free Edition provides, combine with AWS/Azure/GCP student credits for a paid Databricks workspace.
- **Compare with alternatives:** Google Colab (free GPU notebooks), Kaggle (free GPU/TPU), AWS SageMaker Studio Lab (free ML environment). Databricks stands out for its unified data + AI platform.
- **The $100M investment** is spread across curriculum development, free tools, and certification programs globally.

**Sources:**
- https://www.databricks.com/university
- https://www.databricks.com/resources/student-resources
- https://www.databricks.com/company/newsroom/press-releases/databricks-launches-free-edition-and-announces-100-million
- https://docs.databricks.com/aws/en/getting-started/free-edition-limitations
- https://community.databricks.com/t5/databricks-university-alliance/free-edition-frequently-asked-questions-faqs-consolitated/ta-p/128500

## Submission 87: Cursor Pro Free for Students (1 Year)

- Category: Education & Student
- Source Link: https://cursor.com/students
- Amount: 1 year of Cursor Pro (valued $240/year) with 500 fast premium requests/month
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Cursor offers verified students **1 full year of Cursor Pro** completely free — the same $20/month plan that professional developers pay for, a **$240 value**. You get **500 fast premium requests per month** with access to top AI models (Claude, GPT, Gemini), plus unlimited slow requests. Verified through SheerID with a .edu email or enrollment proof. Available in 40+ countries. Announced May 2025.

---

## Registration (Step-by-Step)

1. Go to [cursor.com/students](https://cursor.com/students)
2. **Create a Cursor account** or sign in with an existing one
3. Click **"Get Student Plan"** or equivalent button
4. You will be redirected to **SheerID** for verification
5. Select your **country** and **institution** from the dropdown
6. Enter your **.edu email** (or equivalent academic email for non-US institutions)
7. **Instant verification:** If your institution is recognized and email matches, you are verified in minutes
8. **Document upload:** If instant verification fails, upload enrollment proof:
   - Student ID card (current academic year)
   - Class schedule
   - Enrollment letter
   - Tuition receipt
9. **Add a payment method** (credit card or debit card) — you will NOT be charged during the free year
10. Your Cursor Pro subscription activates immediately

**Common issues and fixes:**
- **"Institution not found"** — try alternate spellings, abbreviations, or the full official name
- **Verification fails** — make sure the email on your Cursor account matches what you submit to SheerID
- **Country not listed** — the program is only available in ~40 countries (see list below). Many countries are still not supported
- **Disable VPNs** during verification — location checks can fail with VPNs active

---

## What You Get

| Feature | Details |
|---|---|
| **Fast Premium Requests** | 500/month — used for AI code completions, chat, and agent mode |
| **Slow Premium Requests** | Unlimited — queued requests that run at lower priority |
| **AI Models** | Claude (Sonnet 4.5, Opus 4.6), GPT (4.1, 5 mini, 5), Gemini (3 Pro, Flash), Cursor's own model |
| **Tab Completion** | AI-powered multi-line code completions |
| **Cursor Chat** | In-editor AI chat for questions, debugging, explanations |
| **Composer** | Multi-file AI editing and code generation |
| **Agent Mode** | Autonomous task execution across your codebase |
| **Codebase Context** | AI understands your full project structure |
| **Terminal Integration** | AI-powered terminal commands |

This is the **full Pro plan** — identical to what paying users get. No features are restricted for students.

---

## Eligibility & Verification

- **Age:** 13+ years old
- **Status:** High school students, college/university students, bootcamp participants
- **Verification:** SheerID (instant via academic email or manual document upload)
- **Email requirement:** .edu or equivalent institutional email

### Supported Countries (~40)

Australia, Austria, Belgium, Bulgaria, Canada, Chile, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Georgia, Germany, Greece, Hungary, Ireland, Israel, Italy, Japan, Latvia, Lithuania, Netherlands, New Zealand, Norway, Poland, Portugal, Romania, Serbia, Slovakia, Slovenia, South Africa, Spain, Sweden, Switzerland, United Arab Emirates, United Kingdom, United States

**Notable exclusions:** India, China, Brazil, South Korea, most of Southeast Asia, most of Africa and Latin America. There are active community requests to expand to more countries.

---

## Auto-Renewal & Cancellation

- After 12 months, the subscription **auto-renews at $20/month**
- Cancel before the trial ends to avoid charges
- **Set a calendar reminder** 2 weeks before your free year expires
- After cancellation, you revert to Cursor's free plan (limited completions per month)

---

## Additional Tips

- **500 fast requests/month is generous** — most student projects will not exhaust this. Each request to a model like Claude Sonnet or GPT-4.1 counts as 1 fast request
- **Use slow requests for non-urgent tasks** — they are unlimited and use the same models, just with a queue delay
- **Cursor vs GitHub Copilot:** Both are free for students. Cursor gives you more model choice and a more integrated agent/composer experience. Copilot integrates with the broader GitHub ecosystem. Many devs use both
- **Note the country restriction** — this is the biggest limitation. If your country is not listed, you cannot get the student discount regardless of your student status
- **Bootcamp students qualify** — this is broader than many student programs that only cover degree-granting institutions
- **High school students qualify** — most other AI student programs require age 18+ or university enrollment. Cursor requires only age 13+
- **Bug reports:** Some students have reported issues with the student verification or plan activation. Check the [Cursor Community Forum](https://forum.cursor.com/) for troubleshooting threads
- **Related programs:** Also check Windsurf (free student plan), GitHub Copilot (free for students), and Cody by Sourcegraph (free tier) for complementary AI coding tools

---

**Sources:**
- [Cursor Students — Official Page](https://cursor.com/students)
- [Cursor Docs — Models](https://cursor.com/docs/models)
- [Sealos Blog — Cursor Free for Students](https://sealos.io/blog/cursor-free-for-students)
- [FastGPT Plus — Cursor Student Discount Guide](https://fastgptplus.com/en/posts/cursor-student-discount-guide)
- [AlternativeTo — Cursor Pro Free for Students](https://alternativeto.net/news/2025/5/cursor-makes-its-pro-ai-coding-plan-free-for-a-year-for-verified-students-worldwide/)
- [Cursor Community Forum — Student Discount Thread](https://forum.cursor.com/t/student-discount-details-updates-q-as/88907)

## Submission 86: Coze Free Plan (ByteDance)

- Category: AI API Free Tiers
- Source Link: https://www.coze.com/
- Amount: 10 daily message credits (~100 GPT-3.5 messages or ~5 GPT-4o messages per day)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

ByteDance's **Coze** is a no-code AI agent/chatbot building platform with a free plan that includes **10 daily message credits** and access to models like GPT-3.5, GPT-4o, and Gemini. You can build chatbots and AI agents using a visual builder with **60+ plugins**, then deploy them to **Discord, Telegram, Slack**, and other platforms. The free plan is restrictive on advanced models (only 1-5 messages/day for GPT-4 class models), but GPT-3.5 gets up to 100 messages daily. Coze has been progressively reducing free limits, so grab it while it lasts.

---

## Registration (Step-by-Step)

1. Go to [coze.com](https://www.coze.com)
2. Click **Sign Up** and create an account (email or social login)
3. You are automatically on the **Free plan** with 10 daily message credits
4. Open the **Bot Builder** to start creating your first AI agent
5. Choose your base model (GPT-3.5 for maximum messages, GPT-4o for quality)
6. Add plugins, knowledge bases, and workflows as needed
7. Deploy your bot to your preferred platform (Discord, Telegram, Slack, etc.)

**Important:**
- Free plan requires no credit card
- Daily message credits reset each day and do not accumulate
- Different models consume different amounts of credits per message
- Coze has been reducing free limits over time — previously offered 50 messages/day, now down to 10 credits

---

## Available AI Models

| Model | Credits per Message | Free Plan Daily Limit |
|---|---|---|
| **GPT-3.5** | 0.1 credits | ~100 messages |
| **Gemini 1.5 Flash** | 0.5 credits | ~20 messages |
| **GPT-4o (basic, 8K)** | 2 credits | ~5 messages |
| **Gemini 1.5 Pro** | 2.5 credits | ~4 messages |
| **GPT-4o (advanced)** | 5 credits | ~2 messages |
| **GPT-4 Turbo** | 10 credits | ~1 message |
| **GPT-4o 128K** | 10 credits | ~1 message |

---

## Free Tier Limits

| Feature | Free Plan | Premium ($9/mo) | Premium Plus ($39/mo) |
|---|---|---|---|
| Daily message credits | 10 | 100 | 1,000 |
| GPT-3.5 messages/day | ~100 | Unlimited | Unlimited |
| GPT-4o basic messages/day | ~5 | 50 | 500 |
| GPT-4 Turbo messages/day | ~1 | 10 | 100 |
| Bot creation | Yes | Yes | Yes |
| Plugin access (60+) | Yes | Yes | Yes |
| Deploy to platforms | Yes | Yes | Yes |
| Trial period | — | 3 days free | 3 days free |

---

## Platform Features

- **No-code bot builder** — drag-and-drop interface for creating AI chatbots and agents
- **60+ plugins** — extend your bots with tools for web search, image generation, data analysis, and more
- **Knowledge base** — upload documents and data for your bot to reference (RAG)
- **Workflows** — create multi-step automation flows
- **Multi-platform deployment** — publish bots to Discord, Telegram, Slack, WhatsApp, and web embed
- **API & Web SDK** — programmatic access for building custom integrations
- **Coze Tokens** — billing unit for API and SDK usage (separate from daily message credits)

---

## Additional Tips

- **GPT-3.5 is your best bet on free** — at 0.1 credits per message, you get ~100 messages/day, far more than any other model
- **Gemini 1.5 Flash** is the best mid-range option at 0.5 credits (~20 messages/day) with good reasoning capability
- **Free limits have been shrinking** — Coze originally offered free GPT-4 access with 50 messages/day, then cut to 15, then to 1. The trend suggests further restrictions are possible
- **Premium trials** — both paid plans offer 3-day free trials, useful for testing whether upgraded limits are worth it
- **Deploy to Discord for free** — one of the easiest ways to get a custom AI bot running on Discord without any hosting costs
- **Knowledge base RAG** — upload your docs to create a domain-specific assistant at no extra cost on the free plan
- **API access** uses Coze Tokens (a separate billing system) — the free plan's daily message credits apply to the web interface and deployed bots
- **ByteDance backing** — the platform is well-funded but the long-term pricing strategy is still evolving

---

**Sources:**
- [Coze Platform](https://www.coze.com)
- [Coze Premium Plans](https://www.coze.com/premium)
- [Coze Billing Documentation](https://www.coze.com/docs/guides/billing)
- [Coze Message Credits Guide](https://www.coze.com/open/docs/guides/message_credits)
- [Coze Premium Plans Guide](https://www.coze.com/docs/guides/subscription?_lang=en)
- [ByteDance Coze Pricing Analysis (Oreate AI)](https://www.oreateai.com/blog/indepth-analysis-of-bytedances-coze-platform-paid-plans-pricing-strategy-and-market-positioning-assessment/76c7269cbaebdc600588687b1f25eea9)

## Submission 85: Cohere For AI Scholars Program (Paid Research Fellowship)

- Category: Education & Student
- Source Link: https://cohere.com/research/scholars-program
- Amount: Competitive full-time salary + large-scale compute access + mentorship for 8 months
- Difficulty: 5/5 (Very Hard)
- Expires At: N/A

### Description

The Cohere For AI (now Cohere Labs) Scholars Program is an **8-month paid, remote-first research fellowship** for early-career ML researchers. Scholars receive a competitive salary, mentorship from Cohere's research team, and access to large-scale compute infrastructure. The program is designed to open doors for emerging researchers globally, with a strong emphasis on diversity -- Cohere actively recruits from underrepresented regions like Latin America and Africa. No prior publications are required; curiosity and engineering skills matter more. The third cohort opened applications in 2025, and the program typically runs January through August.

---

## Registration (Step-by-Step)

1. **Watch for the application window:** Applications typically open in **August** each year. Follow [Cohere Labs on X/Twitter](https://x.com/Cohere_Labs) and [LinkedIn](https://www.linkedin.com/company/cohere-labs/) for announcements
2. **Visit the program page:** [cohere.com/research/scholars-program](https://cohere.com/research/scholars-program)
3. **Prepare your application:** You will likely need:
   - Your CV/resume
   - A statement of research interests
   - Description of relevant ML/engineering experience (publications not required)
   - Evidence of demonstrated AI/ML curiosity (projects, contributions, coursework)
4. **Submit your application** through the online portal during the open window
5. **Selection process:** Competitive review; successful applicants are matched to projects based on their skills, interests, and background
6. **Program start:** Selected scholars begin in **January** of the following year
7. **Commitment:** Full-time, 8-month engagement (January through August)

---

## What You Get

### Compensation & Resources

| Benefit | Details |
|---|---|
| **Salary** | Competitive, industry-level pay (full-time position) |
| **Compute access** | Large-scale experimental infrastructure for ML research |
| **Mentorship** | Personalized 1-on-1 mentorship from Cohere's ML research and engineering experts |
| **Research projects** | Matched to projects based on your skills and interests |
| **Publication support** | Scholars are supported to work toward publication within the program period |
| **Remote-first** | No relocation required; work from anywhere globally |
| **Office access** | Optional in-person access at Cohere offices for those who want it |

### Research & Career Development

- **Lead research projects** -- scholars are not just assistants; they are key contributors positioned to drive research
- **Explore complex ML problems** with cutting-edge compute that would be difficult to access independently
- **Build your network** within Cohere Labs and the broader ML research community
- **Publication track** -- the program is structured to produce publishable research within the 8-month timeframe
- **Career pipeline** -- strong performers may receive offers for full-time research positions at Cohere

---

## Eligibility & Verification

- **Target audience:** Early-career ML researchers (PhD students, recent graduates, self-taught researchers)
- **Full-time availability required** for the 8-month program duration
- **Prior publications are NOT required** -- Cohere explicitly states this to lower barriers
- **What they look for:**
  - Demonstrated AI/ML curiosity
  - Strong engineering skills
  - Ability to work independently on research problems
  - Passion for advancing ML research
- **Global eligibility:** Remote-first, no geographic restrictions. Cohere particularly encourages applicants from Latin America and Africa
- **No specific degree requirement mentioned** -- practical skills and curiosity are emphasized over credentials

---

## Additional Tips

- **Applications are highly competitive.** The program receives many applicants globally. Make your application stand out by clearly articulating your research interests and demonstrating concrete ML engineering skills (GitHub projects, Kaggle competitions, open-source contributions)
- **No publications needed, but they help.** While Cohere explicitly says publications are not required, having preprints, workshop papers, or open-source ML projects will strengthen your application
- **Remote-first is real.** Previous scholars have participated from around the world. You do not need to be in North America
- **Diversity focus:** Cohere has publicly stated they want more applicants from LatAm and Africa. If you are from an underrepresented region, you are strongly encouraged to apply
- **This is a full-time paid position**, not a part-time internship or volunteer opportunity. You will need to commit 40 hours/week for 8 months. If you are a PhD student, discuss with your advisor whether this fits your program timeline
- **Timeline summary:**
  - **August:** Applications open
  - **September-November:** Review and selection
  - **January:** Program starts
  - **August:** Program ends (8 months)
- **Alternative programs:** If you miss the window or are not selected, consider Google DeepMind internships, Microsoft Research internships, Meta AI residency, or the NVIDIA Graduate Fellowship as other research opportunities
- **Cohere Labs vs. Cohere:** The Scholars Program is run by Cohere Labs (formerly Cohere For AI), the nonprofit research arm. It focuses on open research, not product development

---

**Sources:**
- [Cohere Labs Scholars Program](https://cohere.com/research/scholars-program)
- [Cohere For AI Scholars Program 2025 -- Blog](https://cohere.com/blog/cohere-for-ai-scholars-program-2025)
- [Your Research Journey Starts Here -- Blog](https://cohere.com/blog/c4ai-scholars-program)
- [Introducing the Scholars Program -- Blog](https://cohere.com/blog/introducing-the-cohere-for-ai-scholars-program-your-research-journey-starts-here)
- [Cohere Labs on X/Twitter](https://x.com/Cohere_Labs)

## Submission 84: Cohere Catalyst Grants (Free API Credits for Research)

- Category: Education & Student
- Source Link: https://cohere.com/research/grants
- Amount: Free Cohere API credits (amount varies by project scope and needs)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Cohere Labs (formerly Cohere For AI) provides free API credits to academic researchers, civic institutions, and public-benefit organizations through their Catalyst Grants program. The credits give you production-level access to Cohere's full model suite -- Command, Aya, Embed, and Rerank -- so you can run real research without worrying about API costs.

## Registration (Step-by-Step)

1. Go to [cohere.com/research/grants](https://cohere.com/research/grants) and review the program details.
2. Click through to the [application form](https://cohere.com/research/grants/application).
3. Fill out the application with:
   - Your name and affiliation (university, research lab, or organization).
   - Project description explaining what you're building/researching.
   - Expected API usage and which models you need.
   - How the project serves public benefit or advances open science.
4. Submit and wait -- applications are reviewed on a **rolling basis** (no fixed deadline).
5. If approved, you'll receive Production API keys with credits loaded to your Cohere dashboard.
6. Monitor your credit usage in the Cohere dashboard under the billing tab.
7. You can invite collaborators to your organization to share the credits.

## What You Get

- **Production API access** to all Cohere models:
  - **Command R / Command R+** -- instruction-following LLMs (128K context), great for text generation, summarization, RAG.
  - **Aya Expanse (8B & 32B)** -- multilingual models covering 23+ languages.
  - **Aya Vision** -- multimodal model for text + image tasks.
  - **Embed 4** -- multimodal embedding model for semantic search and retrieval.
  - **Rerank 3.5** -- re-ranking model to boost search relevance.
- **Credit amount varies by project** -- Cohere tailors the grant to your stated needs.
- **Additional credits available** -- you can request more if your project scales.

## Eligibility & Verification

- **Academic researchers** at universities or research institutions.
- **Civic/public-benefit organizations** (nonprofits, NGOs, government research bodies).
- Projects must focus on **public benefit or open science**.
- Priority research areas:
  - AI safety (bias, explainability, hallucinations, toxicity, adversarial robustness).
  - Prosocial applications (education, healthcare, climate science, legal tech, social science).
  - Language model capabilities and efficiency.
  - Multilingual AI beyond English.
  - Values alignment.
- **AI safety research disclosure requirement:** If your work involves safety research, you must share findings with safety@cohere.ai at least 4 weeks before publication.

## Additional Tips

- **Rolling applications** -- no deadline, but be specific about your project scope and expected usage. Vague proposals likely get rejected.
- **Free trial exists for everyone** -- even without a Catalyst Grant, Cohere offers a free Trial API key with rate-limited access to all endpoints. Good for prototyping before applying.
- **Cohere's free tier pricing** -- Trial keys are rate-limited but free. If you outgrow the trial but don't qualify for Catalyst, check their pay-as-you-go pricing (Aya Expanse starts at $0.50/1M input tokens).
- **Pair with other programs** -- combine Catalyst credits with cloud credits from AWS, GCP, or Azure for a fully free research stack.
- **Cohere Labs community** -- join their Discord and community forums for support and to connect with other researchers.

**Sources:**
- https://cohere.com/research/grants
- https://cohere.com/research/grants/application
- https://cohere.com/blog/granting-access
- https://docs.cohere.com/docs/models
- https://cohere.com/pricing

## Submission 83: Codesphere Free GPU Workspace

- Category: Free GPU & Compute
- Source Link: https://codesphere.com/
- Amount: 1 free GPU workspace with 20 GB storage (shared GPU, unlimited hours but auto-shutdown after 60 min idle)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Codesphere offers **one free GPU workspace** per user with 20 GB of storage and a shared GPU — no credit card required. Unlike notebook-only platforms (Colab, Kaggle), Codesphere gives you a full cloud development environment where you can run any workload, not just Jupyter notebooks. The free tier includes 3 workspaces (with GPU on one) and 3 team seats. The catch: free workspaces auto-shutdown after ~60 minutes of inactivity, and GPU availability is limited by demand.

---

## Registration (Step-by-Step)

1. Go to [codesphere.com](https://codesphere.com/) or the early access signup at [signup.codesphere.com](https://signup.codesphere.com)
2. Create an account — no credit card required
3. You receive **3 free workspaces** and **1 free shared GPU** automatically
4. Create a new workspace and select the GPU option
5. Choose from available templates (Stable Diffusion, LLM inference, etc.) or start with a blank environment
6. Start coding — your workspace includes a full IDE, terminal, and deployment pipeline

**Caveats:**
- Free workspaces **turn off after ~60 minutes of inactivity** (no IDE activity). They boot back up in seconds when you return
- GPU availability is constrained — Codesphere has stated they sometimes have more demand than capacity. If you can't get a GPU workspace, try again later or upgrade to paid
- The GPU is **shared**, not dedicated — expect variable performance
- Only your initial team gets the 3 free workspace allocation. Additional teams require paid plans

---

## GPU Hardware & Specs

| Feature | Detail |
|---|---|
| Free GPU type | Shared GPU (specific model not publicly disclosed) |
| Paid GPU options | NVIDIA T4, A100 (available on-demand) |
| Free storage | 20 GB |
| Paid GPU add-on | ~$10/month per GPU with "off when unused" feature |

The free shared GPU is suitable for:
- Running smaller AI models and inference
- Stable Diffusion image generation
- Fine-tuning small models
- General GPU-accelerated development

For heavy training workloads, you would need to upgrade to paid T4 or A100 instances.

---

## Free Tier Limits

| Feature | Free Tier | Paid (Micro $5/mo) | Paid (Boost $20/mo) |
|---|---|---|---|
| Workspaces | 3 | More | More |
| Storage | 20 GB | 1 GB (base) | More |
| GPU | 1 shared GPU | Add-on available | Add-on available |
| Team seats | 3 free | +$10/seat/mo | +$10/seat/mo |
| Auto-shutdown | ~60 min idle | Configurable | Configurable |
| Credit card | Not required | Required | Required |

---

## Platform Features

- **Full cloud IDE** — not just notebooks. You get a VS Code-like editor, terminal, file explorer, and Git integration
- **One-click deployment** — deploy your workspace as a live web service with a URL
- **Docker support** — run any containerized workload
- **Collaboration** — share workspaces with your team in real time
- **Managed PostgreSQL** — available as an add-on service
- **Templates** — pre-built templates for common AI workloads (Stable Diffusion, LLMs, etc.)

---

## Additional Tips

- **Not just for notebooks** — the biggest differentiator vs. Google Colab or Kaggle is that Codesphere gives you a full dev environment. You can build and deploy full applications, not just run experiments
- **Auto-shutdown workaround** — the 60-minute idle timeout applies to IDE activity specifically. A planned "Dev Deployment mode" was announced that would keep workspaces alive when users visit their dev domain, but check if this has shipped
- **GPU availability is hit-or-miss** — if the free GPU slots are full, you'll need to wait or pay. Try signing up during off-peak hours (nights, weekends) for better luck
- **Good for hobbyists and prototyping** — the free tier is best suited for hobbyist projects, learning ML, and quick prototyping. For serious training, use a dedicated GPU cloud (RunPod, Vast.ai, Lambda)
- **Combine with other free tiers** — use Codesphere for development and deployment, and a free GPU provider (Colab, Kaggle, Lightning AI) for heavy compute tasks
- **$10/month GPU add-on is competitive** — if you like the platform, the paid GPU option with "off when unused" billing is very affordable for intermittent workloads

---

**Sources:**
- [Codesphere Homepage](https://codesphere.com/)
- [Codesphere Free Tier Launch Announcement](https://codesphere.com/articles/free-tier-launch)
- [Codesphere Pricing](https://ai.codesphere.com/pricing)
- [Codesphere — Best Free Cloud GPU Providers for Hobbyists](https://codesphere.com/articles/5-best-free-cloud-gpu-providers-for-hobbyists)
- [Codesphere Pricing & Plans — UpendraVarma](https://upendravarma.com/codesphere-pricing/)

## Submission 82: Cloudflare Workers Paid Plan Free for Students (1 Year)

- Category: Education & Student
- Source Link: https://www.cloudflare.com/students
- Amount: 1 year of Workers Paid plan free (valued $60/year) -- includes 10M Workers requests/month, KV, Durable Objects, Containers, Vectorize, and more
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Cloudflare gives US university students with a .edu email one full year of the Workers Paid plan at no cost (normally $5/month). This includes massively upgraded limits on Workers, KV, Durable Objects, and more compared to the free tier. If you're building serverless apps, APIs, or AI-powered tools on the edge, this is a solid deal.

## Registration (Step-by-Step)

### New Cloudflare Account
1. Go to [cloudflare.com/students](https://www.cloudflare.com/students/).
2. Sign up for a Cloudflare account if you don't have one.
3. **Add a credit card** to your account (required for the paid plan, but you won't be charged during the student year).
4. Go to your **billing settings** and add/verify your **.edu email** as the billing email.
5. Navigate to the **Workers & Pages** section in the dashboard.
6. Select the **Workers Paid plan** -- the student discount should apply automatically.

### Existing Cloudflare Account
1. Log in to your Cloudflare dashboard.
2. Go to billing settings and **change your billing email to your .edu address**.
3. Verify the .edu email.
4. Add the Workers Paid plan -- the upfront cost is waived for 12 months.

**Caveat:** A credit card is required even though you won't be charged. This is standard for Cloudflare paid plans.

## What You Get

| Service | Student (Paid) Plan | Free Plan |
|---------|--------------------|-----------|
| **Workers requests** | 10M requests/month | 100K requests/day |
| **Workers KV reads** | 10M reads/month | 100K reads/day |
| **Workers KV writes** | 1M writes/month | 1K writes/day |
| **Durable Objects** | 1M requests/day | Not available |
| **Hyperdrive** | Unlimited DB queries | Not available |
| **Workers Logs** | 20M events/month (7-day retention) | Not available |
| **Workers Logpush** | 10M events/month | Not available |
| **Queues** | 1M operations/month | Not available |
| **Pages Functions** | Included | Limited |
| **Containers** | Included | Not available |
| **Vectorize** | Included | Prototype only |

- **Workers AI** -- available on both free and paid plans, priced at $0.011 per 1,000 Neurons (usage-based, not part of the free student allotment).
- If you exceed the included limits, **standard usage rates apply** (you'd be billed on your credit card).

## Eligibility & Verification

- Must be a **US university or community college student**.
- Must be **18 years or older**.
- Must have a valid **.edu email address**.
- The .edu email must be set as your **billing email** in Cloudflare.
- **US only** -- students outside the US are not eligible at this time.

## Additional Tips

- **After 12 months**, your account automatically reverts to the free plan. No surprise charges -- you just lose the paid-tier limits.
- **You can upgrade** to the regular Workers Paid plan ($5/month) after the student year ends to keep your limits.
- **Workers AI is usage-based** -- it's not fully free even on the student plan. Keep an eye on Neuron consumption if you're running inference workloads.
- **Dedicated student Discord channel** -- Cloudflare runs a Discord community with a channel specifically for students. Good for getting help and networking.
- **Pair with other student perks:**
  - GitHub Student Developer Pack (free Copilot, domains, etc.)
  - Azure for Students ($100 credits)
  - Google Cloud student credits
- **Great for deploying side projects** -- Workers + KV + Durable Objects is enough stack for most indie apps, APIs, and bots.
- **Cloudflare also has a startup program** -- if your student project turns into a business, check out Cloudflare for Startups for continued free/discounted access.

**Sources:**
- https://www.cloudflare.com/students/
- https://blog.cloudflare.com/workers-for-students/
- https://developers.cloudflare.com/workers/platform/pricing/
- https://developers.cloudflare.com/workers-ai/platform/pricing/

## Submission 81: Cerebrium Free Serverless GPU Credits - $10 Free + Up to $1,000 for Startups

- Category: Free GPU & Compute
- Source Link: https://www.cerebrium.ai/
- Amount: $10 free GPU credits (up to $1,000 for qualified startups). 9 GPU types from T4 ($0.59/hr) to H200 ($3.30/hr).
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Cerebrium is a **serverless AI infrastructure platform** (YC-backed) that gives new accounts **$10 in free GPU credits** with access to **9 GPU types from T4 to H200**. The platform features **pay-per-second billing** with no idle costs, cold starts as low as 2-4 seconds, and automatic scaling from zero to thousands of requests. Qualified startups can apply for up to **$1,000 in credits** with dedicated engineering support. No credit card required for the free Hobby tier.

---

## Registration (Step-by-Step)

1. Go to [cerebrium.ai](https://www.cerebrium.ai/) and click **Get Started**
2. Create an account with your email
3. You'll immediately receive **$10 in free GPU credits** on the Hobby plan
4. Install the CLI: `pip install cerebrium`
5. Authenticate: `cerebrium login` and enter your API key
6. Initialize a project: `cerebrium init my-project`
7. Configure GPU in `cerebrium.toml` under `[cerebrium.hardware]` and deploy with `cerebrium deploy`

**For startup credits ($1,000):**
- Apply through Cerebrium's startup program on their website
- Includes face-time with Cerebrium engineers for onboarding
- Must be a qualified startup (criteria not publicly detailed)

---

## GPU Hardware & Specs

| GPU Model | Identifier | VRAM | Price/Hour | Plan Required |
|---|---|---|---|---|
| **NVIDIA T4** | TURING_T4 | 16 GB | $0.59 | Hobby+ |
| **NVIDIA L4** | ADA_L4 | 24 GB | $0.80 | Hobby+ |
| **NVIDIA A10** | AMPERE_A10 | 24 GB | $1.10 | Hobby+ |
| **NVIDIA A100 (40GB)** | AMPERE_A100_40GB | 40 GB | $1.45 | Enterprise |
| **NVIDIA L40S** | ADA_L40 | 48 GB | $1.95 | Hobby+ |
| **NVIDIA A100 (80GB)** | AMPERE_A100_80GB | 80 GB | $2.06 | Enterprise |
| **NVIDIA H100** | HOPPER_H100 | 80 GB | $2.21 | Enterprise |
| **NVIDIA H200** | - | 141 GB | $3.30 | Enterprise |
| **AWS Inferentia 2** | INF2 | 32 GB | - | Hobby+ |
| **AWS Trainium** | TRN1 | 32 GB | - | Hobby+ |

All GPUs support up to **8x multi-GPU** configurations for distributed workloads.

---

## Free Tier Limits

| Resource | Limit |
|---|---|
| **Free credits** | $10 (new accounts) |
| **Startup credits** | Up to $1,000 (qualified startups) |
| **Hobby plan cost** | $0/month |
| **GPUs on Hobby plan** | T4, L4, A10, L40S, Inferentia, Trainium |
| **GPUs requiring Enterprise** | A100, H100, H200 |
| **Block storage (free)** | First 100 GB included |
| **Additional storage** | $5/month per additional block |
| **Cold start time** | 2-4 seconds typical |
| **Billing model** | Pay-per-second, no idle costs |
| **Credit card required** | No (Hobby plan) |

---

## Plan Tiers

| Feature | Hobby ($0/mo) | Standard ($100/mo) | Enterprise (Custom) |
|---|---|---|---|
| **Free credits** | $10 | - | Up to $1,000 |
| **GPU access** | T4, L4, A10, L40S | All standard GPUs | All GPUs including H100, A100, H200 |
| **Multi-GPU** | Up to 8x | Up to 8x | Up to 8x |
| **Storage** | 100 GB free | Included | Custom |
| **Support** | Community | Priority | Dedicated engineering |

---

## Configuration Example

```toml
# cerebrium.toml
[cerebrium.hardware]
compute = "AMPERE_A10"
gpu_count = 1
cpu = 4
memory = 16
```

Or via CLI:
```bash
cerebrium deploy --compute AMPERE_A10 --gpu-count 1
```

---

## Additional Tips

- **$10 goes further than you think**: At T4 pricing ($0.59/hour billed per-second), $10 buys about 17 hours of GPU compute. If your inference calls take 5 seconds each, that's roughly 12,000 API calls
- **Hobby plan is genuinely useful**: Unlike many "free tiers" that are crippled, the Hobby plan gives you access to solid GPUs (T4, L4, A10, L40S) and 100 GB storage
- **Cold starts are fast**: 2-4 second cold starts are competitive with Modal and faster than most alternatives. Code at the top level of your module runs only on first container spin-up
- **VRAM calculation**: Use the formula `modelVRAM = parameters x bytes_per_dtype x 1.5` to estimate GPU requirements. The 1.5x buffer accounts for runtime overhead
- **Multi-GPU for large models**: You can run up to 8 GPUs in a single deployment, making it possible to serve 70B+ parameter models with tensor parallelism
- **YC-backed**: Cerebrium is a Y Combinator company, which may give you confidence in platform longevity and may help if you're also a YC company
- **Mistral integration**: Cerebrium has official Mistral AI deployment guides, making it easy to self-host Mistral models
- **Compare with Modal**: Modal offers $30/month in free credits vs Cerebrium's one-time $10. But Cerebrium's per-second GPU pricing is often cheaper for sustained workloads
- **AWS chip support**: Cerebrium uniquely offers AWS Inferentia 2 and Trainium chips on the Hobby plan, which can be significantly cheaper for inference and training respectively

---

**Sources:**
- [Cerebrium Official Site](https://www.cerebrium.ai/)
- [Cerebrium Pricing](https://www.cerebrium.ai/pricing)
- [Cerebrium GPU Documentation](https://docs.cerebrium.ai/cerebrium/hardware/using-gpus)
- [Cerebrium on Y Combinator](https://www.ycombinator.com/companies/cerebrium)
- [Cerebrium Pricing Deep Dive - Skywork AI](https://skywork.ai/skypage/en/Cerebrium-Pricing-A-2025-Deep-Dive-into-Serverless-GPU-Costs/1975270663121465344)
- [Cerebrium Review - GetDeploying](https://getdeploying.com/cerebrium)
- [Cerebrium on Product Hunt](https://www.producthunt.com/products/cerebrium)
- [Free GPU Cloud Trials 2026 - GMI Cloud](https://www.gmicloud.ai/blog/where-can-i-get-free-gpu-cloud-trials-in-2026-a-complete-guide)

## Submission 80: Beam Cloud Free GPU Tier - 10 Hours Serverless GPU

- Category: Free GPU & Compute
- Source Link: https://www.beam.cloud/
- Amount: 10 hours free GPU compute on signup, up to 3 concurrent applications
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Beam Cloud is an **open-source serverless GPU platform** (powered by the [beta9 runtime](https://github.com/beam-cloud/beta9)) that gives you **10 free GPU compute hours on signup**. No credit card required. The platform supports T4, A10G, RTX 4090, A100, and H100 GPUs with **pay-per-second billing**, cold starts of 2-3 seconds, and warm starts as fast as 50ms. Deploy AI models as serverless REST APIs, run async task queues, or launch Jupyter notebooks -- all with instant autoscaling and no idle costs.

---

## Registration (Step-by-Step)

1. Go to [beam.cloud](https://www.beam.cloud/) and click **Get Started**
2. Sign up with your email or GitHub account
3. No credit card is required -- you immediately get **10 hours of free GPU compute**
4. Install the Beam CLI: `pip install beam-client` (or `curl https://raw.githubusercontent.com/beam-cloud/beta9/main/bin/install.sh -sSfL | bash`)
5. Authenticate with `beam login` and enter your API key from the dashboard
6. Write your first function using the `@endpoint` decorator and deploy with `beam deploy`

**Important:**
- The free tier is limited to **3 concurrent applications**
- Free credits are one-time; after exhausting them, you move to pay-per-second billing
- No upfront commitments or long-term contracts

---

## GPU Hardware & Specs

| GPU Type | VRAM | Approx. Price/Hour |
|---|---|---|
| **NVIDIA T4** | 16 GB | $0.15 |
| **NVIDIA A10G** | 24 GB | $0.29 - $1.05 |
| **NVIDIA RTX 4090** | 24 GB | $0.19 - $0.69 |
| **NVIDIA A100 (40-80GB)** | 40-80 GB | $0.63 - $3.50 |
| **NVIDIA H100** | 80 GB | $0.97 - $7.15 |

Pricing varies based on CPU, RAM, and GPU configuration. GPU charges are combined with separate CPU core ($0.19/core/hr) and RAM ($0.02/GB/hr) costs. All billed per-second.

---

## Free Tier Limits

| Resource | Limit |
|---|---|
| **Free GPU hours** | 10 hours (one-time on signup) |
| **Concurrent applications** | Up to 3 |
| **Cold start time** | 2-3 seconds typical |
| **Warm start time** | ~50 ms |
| **File storage** | Included free |
| **Credit card required** | No |
| **Billing model** | Pay-per-second after free credits |

---

## Platform Features

- **Serverless REST APIs**: Deploy any Python function as an auto-scaling endpoint
- **Task queues**: Async job processing with automatic retries
- **Scheduled jobs**: Cron-style scheduling for recurring workloads
- **Jupyter notebooks**: GPU-powered notebook environments in the cloud
- **Custom Docker images**: Bring your own container for full environment control
- **Fast cloud storage**: Integrated storage for model weights and datasets
- **Local debugging**: Test locally, deploy to cloud with the same CLI
- **Multi-region**: Infrastructure in US, Europe, and Asia
- **Multi-GPU support**: Available by request via Slack

---

## Additional Tips

- **Open-source self-hosting**: Beam's core runtime (beta9) is fully open-source on GitHub. You can self-host it on your own GPU cluster for free, avoiding any usage limits entirely
- **GPU fallback ordering**: You can specify multiple GPU types as a priority list (e.g., `["A10G", "T4"]`), and Beam will allocate the first available GPU from your list
- **No vendor lock-in**: The same CLI and code works across Beam's managed cloud, your on-prem infrastructure, and hybrid setups
- **Cold start optimization**: Keep functions warm by setting a minimum replica count (costs money but eliminates cold starts for production workloads)
- **Compare with Modal**: Beam and Modal are direct competitors. Beam's advantage is open-source portability; Modal's advantage is deeper Python integration and a larger free tier ($30/month)
- **Combine free tiers**: Use Beam's 10 free hours for serverless API deployment and pair with Kaggle/Colab for interactive notebook training

---

**Sources:**
- [Beam Cloud Official Site](https://www.beam.cloud/)
- [Beam Cloud Pricing](https://www.beam.cloud/pricing)
- [Beam Cloud GPU Pricing - CloudGPUPrices](https://cloudgpuprices.com/vendors/beam)
- [Beam Cloud GPU Documentation](https://docs.beam.cloud/v2/environment/gpu)
- [Beta9 GitHub Repository](https://github.com/beam-cloud/beta9)
- [Serverless GPU Platforms Comparison - Introl](https://introl.com/blog/serverless-gpu-platforms-runpod-modal-beam-comparison-guide-2025)
- [Top Serverless GPU Clouds 2026 - RunPod](https://www.runpod.io/articles/guides/top-serverless-gpu-clouds)

## Submission 79: Arena Academic Partnerships (Up to $50K/Project)

- Category: Education & Student
- Source Link: https://arena.ai/blog/academic-partnerships-program
- Amount: Up to $50,000 per project + access to Arena datasets
- Difficulty: 4/5 (Hard)
- Expires At: 2026-03-31

### Description

Arena (formerly LMArena, the team behind Chatbot Arena / LMSYS) funds independent academic research in AI evaluation, rankings, and measurement. The program provides **up to $50,000 per project** plus access to Arena datasets. Arena raised $150M at a $1.7B valuation and is investing in advancing the scientific foundations of AI evaluation through independent academic work. This is a research grant, not a startup program -- it is designed for tenure-track faculty at U.S. universities.

---

## Registration (Step-by-Step)

1. **Review eligibility:** You must be a tenure-track faculty member at a U.S.-based university or research institution
2. **Prepare your proposal** covering one or more of the supported research areas (evaluation methodology, measurement validity, human preference learning, safety and alignment)
3. **Submit via the Google Form** at [forms.gle/ASE5irwpRWgWLKMu5](https://docs.google.com/forms/d/e/1FAIpQLSdKwhGge-qb8MFvpwk93GBgU0XKkqbGpb6Q3-l9oilRcD7JXg/viewform)
4. **Limit:** One proposal per Principal Investigator (PI)
5. **Wait for review:** Proposals are reviewed quarterly. Authors of accepted proposals are notified approximately 8 weeks after each submission deadline
6. **First deadline: March 31, 2026** (Q1 submission window)

---

## What You Get

- **Research funding:** Up to $50,000 per project
- **Arena datasets:** Selected projects may receive access to Arena's evaluation datasets (user votes, model comparisons, preference data) under appropriate data protection agreements
- **Research amplification:** Arena may help amplify research findings to broader audiences in coordination with the researcher
- **Full independence:** Arena explicitly states their goal is to "support -- not direct" academic research. Funded projects retain full independence and publish through standard academic channels

### Supported Research Areas

- **Evaluation and ranking methodology** -- new approaches to comparing and ranking AI models
- **Measurement and validity** -- statistical foundations of AI evaluation, reliability of benchmarks
- **Learning from human preference** -- how to better capture and use human feedback for model assessment
- **Safety and alignment** -- evaluation methods for AI safety properties
- Other topics that deepen understanding, propose new methods, or critically examine existing approaches to AI evaluation

---

## Eligibility & Verification

- **Who qualifies:** Tenure-track faculty at U.S.-based universities and research institutions
- **Not open to:** Students, postdocs, industry researchers, or faculty outside the U.S. (as of the initial program launch)
- **One proposal per PI** -- no multi-submission from the same researcher
- **Review process:** Quarterly review by Arena's team, with notifications sent approximately 8 weeks after each deadline
- **Quarterly deadlines:** The program operates on a rolling quarterly cycle. Q1 2026 deadline is March 31, 2026

---

## Additional Tips

- **This is a competitive grant** with limited slots. Make your proposal specific and actionable. Arena is looking for research that advances the science of AI evaluation, not general ML research
- **Arena's unique dataset** (millions of human preference votes from Chatbot Arena) is a major draw. If your research can leverage this data, highlight that in your proposal
- **The program launched February 10, 2026**, so it is brand new. Early applicants may have an advantage as the review process is being established
- **For students and postdocs:** While you cannot apply directly, you could be part of a funded project led by an eligible faculty PI. Talk to your advisor about submitting a proposal
- **Arena context:** Arena (formerly LMArena/LMSYS Chatbot Arena) is the most widely used open platform for AI model evaluation, with over 3 million user votes. Their datasets are highly valued in the research community
- **Related programs:** If you are not eligible for this program, consider Google Cloud Research Credits, NVIDIA Academic Hardware Grant, or Microsoft Research grants as alternative academic funding sources
- **Publication freedom:** Arena explicitly supports open publication through standard academic channels. There are no restrictions on where or how you publish your findings

---

**Sources:**
- [Arena Academic Partnerships -- Blog Post](https://arena.ai/blog/academic-partnerships-program/)
- [Arena Academic Partnerships -- PR Newswire](https://www.prnewswire.com/news-releases/arena-launches-academic-partnerships-program-to-advance-ai-evaluation-research-302682931.html)
- [Application Form (Google Forms)](https://docs.google.com/forms/d/e/1FAIpQLSdKwhGge-qb8MFvpwk93GBgU0XKkqbGpb6Q3-l9oilRcD7JXg/viewform)
- [Arena Series A Announcement](https://arena.ai/blog/series-a/)
- [LMArena is now Arena](https://arena.ai/blog/lmarena-is-now-arena/)

## Submission 78: AMD Developer Cloud - Free MI300X GPU Hours

- Category: Free GPU & Compute
- Source Link: https://www.amd.com/en/developer/resources/cloud-access/amd-developer-cloud.html
- Amount: 25 free hours MI300X GPU (192GB HBM3) + up to 50 bonus hours via ROCm Star cert. Credits expire 10 days from deposit.
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

The AMD Developer Cloud provides **free access to AMD Instinct MI300X GPUs with 192 GB HBM3 memory** -- one of the most powerful accelerators available. New developers get **25 free hours** of compute time, and you can earn up to **50 additional hours** by passing the ROCm Star Application Developer Certificate. The platform runs on DigitalOcean infrastructure, supports single-GPU or 8-GPU (1,536 GB total VRAM) configurations, and comes pre-configured with ROCm, PyTorch, vLLM, and other AI frameworks.

---

## Registration (Step-by-Step)

1. Go to [AMD Developer Cloud](https://www.amd.com/en/developer/resources/cloud-access/amd-developer-cloud.html)
2. Click the link to access the cloud -- you'll be redirected to **devcloud.amd.com** (hosted on DigitalOcean infrastructure)
3. Create a new account or log into an existing DigitalOcean account
4. Complete the **Request Credit** form with your developer profile information
5. AMD reviews applications and grants credits at their discretion to qualified developers
6. Once approved, you'll receive **25 hours of free MI300X compute credits** deposited to your account
7. Credits appear in your dashboard and you can immediately launch GPU instances

**Important:**
- Approval is **not automatic** -- AMD grants credits at their sole discretion
- Credits **expire 10 days from the date of deposit** (unless stated otherwise in your confirmation email)
- The 10-day expiration is aggressive -- plan your work before requesting credits
- For questions or issues, contact: devcloudrequests@amd.com

---

## GPU Hardware & Specs

| Configuration | GPU Memory | Details |
|---|---|---|
| **1x MI300X** | 192 GB HBM3 | Single GPU instance, $1.99/hour retail rate |
| **8x MI300X** | 1,536 GB HBM3 (8 x 192 GB) | Multi-GPU cluster, $15.92/hour retail rate |

| Spec | Value |
|---|---|
| **Architecture** | AMD CDNA 3 |
| **Memory Type** | HBM3 |
| **Memory per GPU** | 192 GB |
| **Memory Bandwidth** | 5.3 TB/s |
| **FP16 Performance** | 1,307 TFLOPS |
| **Host CPU** | Intel Xeon Platinum 8568Y+ (Emerald Rapids) |
| **OS** | Ubuntu 24.04 LTS |

---

## Software Stack

| Component | Version/Details |
|---|---|
| **ROCm** | 6.4.0 (pre-configured) |
| **Quick-start packages** | JAX, Megatron, PyTorch, SGLang, vLLM |
| **JupyterLab** | Available for interactive development |
| **OS** | Ubuntu 24.04 LTS |

---

## Free Tier Limits

| Resource | Limit |
|---|---|
| **Initial free hours** | 25 hours MI300X |
| **Bonus hours (ROCm Star cert)** | Up to 50 additional hours |
| **Maximum free hours** | 75 hours total |
| **Credit expiration** | 10 days from deposit |
| **Application required** | Yes, approval at AMD's discretion |
| **Concurrent instances** | Not publicly documented |

---

## ROCm Star Application Developer Certificate

Earn up to **50 additional free hours** by passing the [ROCm Star certification](https://www.amd.com/en/products/software/rocm/application-developer-certificate.html):

- **Format**: 40-question online test
- **Topics covered**: HIP programming, OpenMP offloading, ROCm software-accelerated libraries, porting CUDA applications to ROCm, application design across diverse hardware
- **Reward**: Up to 50 additional free Developer Cloud hours upon passing
- **Purpose**: Validates your ability to develop, refactor, and port applications for AMD GPU hardware

---

## Additional Tips

- **Plan before requesting credits**: The 10-day expiration is the biggest gotcha. Have your project, code, and data ready before you apply. Don't request credits "just to try it out" unless you can dedicate time immediately
- **192 GB VRAM is massive**: This is enough to run very large models (70B+ parameters) at full precision without quantization. Take advantage of this for workloads that won't fit on consumer GPUs
- **ROCm compatibility**: Most PyTorch code works out of the box on ROCm, but CUDA-specific kernels need porting via HIP. Use `torch.cuda.is_available()` -- ROCm maps to the CUDA API
- **vLLM works great on MI300X**: The pre-installed vLLM package is optimized for MI300X and can serve large language models efficiently
- **Noisy neighbors warning**: AMD engineers have cautioned that the shared infrastructure may have variable performance. Don't use it for benchmarking
- **ROCm version may lag**: The pre-installed ROCm 6.4.0 may not be the latest stable release. Check if newer versions are available if you hit compatibility issues
- **Host CPU is Intel, not AMD**: Despite being AMD's cloud, the host processors are Intel Xeon -- this doesn't affect GPU workloads but is worth noting
- **India-specific program**: AMD announced 100,000 free hours specifically for Indian researchers and startups -- if you qualify, this is a much larger allocation
- **Combine with ROCm Star**: Take the certification first to maximize your free hours to 75 total before the 10-day clock starts

---

**Sources:**
- [AMD Developer Cloud](https://www.amd.com/en/developer/resources/cloud-access/amd-developer-cloud.html)
- [Introducing the AMD Developer Cloud - AMD Blog](https://www.amd.com/en/blogs/2025/introducing-the-amd-developer-cloud.html)
- [ROCm Star Application Developer Certificate](https://www.amd.com/en/products/software/rocm/application-developer-certificate.html)
- [AMD ROCm 7 and Developer Cloud Announcement](https://www.amd.com/en/blogs/2025/enabling-the-future-of-ai-introducing-amd-rocm-7-and-the-amd-developer-cloud.html)
- [AMD Developer Cloud Review - Phoronix](https://www.phoronix.com/review/amd-developer-cloud)
- [AMD 100k Free Hours for Indian Developers](https://www.amd.com/en/blogs/2025/100k-hours-free-developer-cloud-access.html)
- [Complimentary AMD Developer Cloud Access - PyTorch Forum](https://dev-discuss.pytorch.org/t/complimentary-access-to-amd-developer-cloud/3103)

## Submission 77: Amazon AI PhD Fellowship ($68M Program)

- Category: Education & Student
- Source Link: https://www.amazon.science/news/amazon-launches-68-million-ai-phd-fellowship-program
- Amount: Full tuition + stipend + travel + AWS cloud credits (from $24M/year pool) + Amazon scientist mentorship
- Difficulty: 5/5 (Very Hard)
- Expires At: 2027-06-30

### Description

Amazon's $68 million AI PhD Fellowship funds over 100 doctoral students across 9 top US universities over two academic years (2025-2027). Fellows receive full tuition, a stipend, travel funding, AWS cloud credits, and one-on-one mentorship from Amazon senior scientists. This is a highly competitive, university-nominated program -- you can't apply directly to Amazon.

## Registration (Step-by-Step)

1. **Check if your university participates.** The 9 partner schools are: Carnegie Mellon, MIT, Stanford, UC Berkeley, UCLA, UT Austin, UIUC, University of Washington, and Johns Hopkins.
2. **You don't apply to Amazon directly.** Each participating university selects its own fellows based on Amazon's criteria (emphasis on practical AI impact).
3. Contact your university's **CS/AI department or graduate office** to ask about the nomination process.
4. If selected, you'll be paired with an **Amazon senior scientist mentor** for regular guidance meetings.
5. AWS cloud credits are distributed through your university's allocation ($24M/year across all schools).

## What You Get

- **Full tuition coverage** for the fellowship period.
- **Living stipend** to cover expenses.
- **Fees covered** (registration, student fees, etc.).
- **AWS cloud computing credits** -- drawn from a pool of $24 million/year distributed across the 9 universities.
- **Amazon mentor** -- each fellow is paired with a senior Amazon scientist for regular meetings on real-world research implications.
- **Travel funding** for conferences and research collaboration.
- **Each university also receives ~$1.1M/year** to support the broader AI research environment.

## Eligibility & Verification

- Must be a **PhD student** at one of the 9 partner universities.
- Research must focus on one of these areas:
  - Machine learning
  - Computer vision
  - Natural language processing
  - Agentic systems
  - Large language models and generative AI
  - Automated reasoning
- Selection is done by the **university**, not by Amazon -- emphasis is on candidates whose research has practical AI impact.
- The first cohort started in the **2025-2026 academic year**.

## Additional Tips

- **This is not an open application.** If you're not at one of the 9 schools, you're not eligible. Consider transferring or collaborating with researchers at partner institutions.
- **AWS credits are shared** across your university's fellows -- the per-student amount depends on how many fellows your school selects.
- **Mentorship is the real value** -- access to Amazon scientists working on production-scale AI can shape your research direction and open career doors.
- **Compare with similar programs:**
  - NVIDIA Graduate Fellowship ($50K stipend + GPU access)
  - Google PhD Fellowship (tuition + stipend + Google mentor)
  - Microsoft Research PhD Fellowship
  - Meta PhD Fellowship
- **Two-year window** -- the program runs for 2025-2026 and 2026-2027 academic years. It's unclear if Amazon will extend it beyond that.
- **15 fellows at UCLA alone** were selected for the first cohort, suggesting roughly 10-15 fellows per university.

**Sources:**
- https://www.amazon.science/news/amazon-launches-68-million-ai-phd-fellowship-program
- https://www.washington.edu/news/2025/10/21/amazon-launches-ai-phd-fellowship-program-with-uw/
- https://hub.jhu.edu/2025/11/03/amazon-ai-phd-fellowship-program/
- https://samueli.ucla.edu/15-ucla-engineering-doctoral-students-named-amazon-ai-fellows/
- https://ai2ai.engineering.jhu.edu/opportunities/fellowship-for-phd-students/

## Submission 76: AI21 Labs $10 Free Trial Credits

- Category: AI API Free Tiers
- Source Link: https://www.ai21.com/pricing
- Amount: $10 free credits, valid 3 months. 200 RPM rate limit.
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

AI21 Labs hands you $10 in free credits to try their Jamba family of language models -- hybrid Mamba-Transformer architecture with 256K context windows. No credit card required. Credits are valid for 3 months. Jamba models are particularly strong at long-context tasks and are ~30% more token-efficient than competitors.

## Registration (Step-by-Step)

1. Go to [studio.ai21.com](https://studio.ai21.com) and click **Sign Up**.
2. Create an account with your email (or sign in via Google/GitHub).
3. Verify your email address.
4. You'll automatically receive **$10 in trial credits** valid for 3 months.
5. Navigate to **API Keys** in the dashboard and generate your key.
6. Use the API at `https://api.ai21.com/studio/v1/` with your key in the `Authorization: Bearer` header.
7. You can also test models in the **AI21 Playground** directly in the browser.

**Caveat:** After 3 months or $10, you'll need to add billing info to continue. Credits do not renew.

## Available AI Models / Services

| Model | Parameters | Active Params | Context | Input Price | Output Price |
|---|---|---|---|---|---|
| Jamba Large (v1.7) | 398B | 94B | 256K tokens | $2.00/M tokens | $8.00/M tokens |
| Jamba Mini (v2) | 52B | 12B | 256K tokens | $0.20/M tokens | $0.40/M tokens |
| Jamba 3B (v2) | 3B | 3B | 256K tokens | Very low | Very low |

All models support 9 languages: English, Spanish, French, Portuguese, Italian, Dutch, German, Arabic, and Hebrew.

AI21 claims ~30% more text per token vs competitors (larger vocabulary), which effectively reduces your cost.

## Free Tier Limits

| Resource | Limit |
|---|---|
| Trial credits | $10 |
| Validity | 3 months from signup |
| Rate limit (RPM) | 200 requests/minute |
| Rate limit (RPS) | 10 requests/second |
| Context window | 256K tokens |
| Seats | Unlimited |

## Additional Tips

- **Best value model:** Jamba Mini at $0.20/M input tokens gives you roughly 50M input tokens with your $10 credit -- that's a lot of experimentation.
- **256K context is the real selling point:** Jamba has one of the longest context windows among commercial models. Great for RAG, document analysis, and summarization.
- **Also available on cloud marketplaces:** Jamba models are available on AWS Bedrock, AWS SageMaker, and Azure -- but those have separate pricing (your AI21 credits don't apply there).
- **Google Startups perk:** Through Google for Startups, you can get up to $10K in credits for Jamba via Vertex AI Model Garden.
- **Token efficiency:** AI21's tokenizer covers more text per token, so direct $/M token comparisons with OpenAI or Anthropic understate the savings.
- **Hybrid architecture:** Jamba's Mamba-Transformer design gives better throughput on long contexts compared to pure Transformer models.

**Sources:**
- https://www.ai21.com/pricing/
- https://docs.ai21.com/docs/jamba-foundation-models
- https://docs.ai21.com/docs/usage-cost
- https://www.ai21.com/jamba/

## Submission 74: xAI Grok API - $150/Month Data Sharing Credits

- Category: AI API Free Tiers
- Source Link: https://console.x.ai/
- Amount: $150/month (with data sharing opt-in) + $25 one-time signup credit
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

xAI offers **$25 in free promotional credits** to every new API account, plus an optional **$150/month in recurring credits** for teams that opt into irrevocable data sharing. The API gives access to the full Grok model family -- from the budget Grok 3 Mini ($0.10/M input) to the flagship Grok 4 ($3/M input) -- with OpenAI- and Anthropic-compatible SDKs, so switching is straightforward. No X Premium subscription is required for API access; you just need an xAI account and credits.

---

## Registration (Step-by-Step)

1. Go to [console.x.ai](https://console.x.ai) or [accounts.x.ai/sign-up](https://accounts.x.ai/sign-up)
2. Click **"Sign Up"** -- you can register with email + password or use **"Login with X"** to authenticate via your X account
3. **Verify your email** by clicking the confirmation link sent to your inbox
4. Log in to the xAI dashboard -- your **$25 promotional credit** should appear within minutes
5. Navigate to **"API Keys"** in the left sidebar and click **"Create New Key"**
6. Copy and securely store the generated key (starts with `xai-`). It is only shown once
7. Authenticate requests with `Authorization: Bearer <your xAI API key>` against the base URL `https://api.x.ai`

**Important:**
- No credit card is required to create an account and use the $25 promotional credit
- Promotional credits expire **30 days** after account creation -- use them or lose them
- The API is OpenAI SDK-compatible: set `base_url="https://api.x.ai/v1"` in the OpenAI Python client and it works
- The Anthropic SDK is also supported for compatible endpoints

---

## Data Sharing Program ($150/Month Credits)

The data sharing program is the main free credit offering. In exchange for letting xAI use your API inputs and outputs to train future models, you get **$150/month in credits** that work with all Grok models.

### How to Enable

1. Your team must have at least **$5 in total prior API spend** (promotional credits count toward this)
2. Go to the **Credits Section** on the xAI Console Billing page
3. Read and accept the data sharing terms
4. Only **team admins** can enable data sharing
5. Credits appear within 24 hours and refresh monthly

### Critical Caveats

- **Irrevocable:** Once a team opts in, it **cannot opt out**. This is permanent for that team
- **Workaround:** You can create multiple teams -- opt one into data sharing for non-sensitive work and keep another team without data sharing for proprietary projects
- **Data shared:** Your prompts and model responses are used by xAI to improve future model versions
- **Not suitable for:** Proprietary code, confidential business data, personal information, or anything you would not want used in AI training
- **Credits do not accumulate** -- unused credits from one month do not roll over
- **Geographic restrictions:** Only available in eligible countries (xAI has not published a definitive list, but major markets in North America, Europe, and parts of Asia are generally supported)

**Note:** There have been reports (May 2025) that xAI ended the data sharing program. However, third-party trackers still list it as active. Check the xAI console directly at [console.x.ai](https://console.x.ai) to confirm current availability before relying on this credit.

---

## Available Models & Pricing

| Model | Input (per 1M tokens) | Output (per 1M tokens) | Context Window | Best For |
|---|---|---|---|---|
| **Grok 4** | $3.00 | $15.00 | 256K | Complex reasoning, flagship |
| **Grok 4.1 Fast** | $0.20 | $0.50 | 2M | High-volume production, best value |
| **Grok 3** | $2.00 | $10.00 | 128K | General tasks |
| **Grok 3 Mini** | $0.10 | $0.30 | 128K | Simple tasks, lowest cost |

### Cached Token Pricing

| Model | Cached Input (per 1M) | Savings |
|---|---|---|
| Grok 4.1 Fast | $0.10 | 50% |
| Grok 4 | $0.75 | 75% |
| Grok 3 | $0.50 | 75% |

### Additional Models

- **Grok 3 Mini** (`grok-3-mini-beta`, `grok-3-mini-fast-beta`) -- supports `reasoning_effort` parameter (low/high)
- **Grok 2** (`grok-2-1212`) -- legacy, still available
- **Grok 2 Vision** (`grok-2-vision-1212`) -- image understanding
- **Grok 2 Image** (`grok-2-image-1212`) -- text-to-image generation
- **Grok Imagine Image** (`grok-imagine-image`) -- image generation
- **Grok Code Fast** (`grok-code-fast-1`) -- optimized for agentic coding tasks

### Tool Pricing

| Tool | Cost |
|---|---|
| Web Search | $5.00 per 1,000 calls |
| X Search | $5.00 per 1,000 calls |
| Code Execution | $5.00 per 1,000 calls |
| Document Search | $5.00 per 1,000 calls |
| Collections Search | $2.50 per 1,000 calls |
| Live Search (per source) | $25.00 per 1,000 sources |

---

## Maximizing Your Credits

With **$150/month** in data sharing credits, here is what you can get:

| Model | Approximate Usage per $150 |
|---|---|
| Grok 4.1 Fast (input) | ~750M input tokens |
| Grok 4.1 Fast (output) | ~300M output tokens |
| Grok 3 Mini (input) | ~1.5B input tokens |
| Grok 4 (input) | ~50M input tokens |

For most indie developers and experimenters, Grok 4.1 Fast offers the best balance of capability and cost -- near-frontier intelligence at 15x less than Grok 4.

---

## Grok Access via X (Non-API)

Separate from the API, Grok is also available as a conversational assistant through the X platform:

| Tier | Price | Grok Access |
|---|---|---|
| **Free (X users)** | $0 | Grok 3, ~10 requests/2 hours, Aurora image gen |
| **X Premium+** | ~$16/month | Full Grok 4, higher limits |
| **SuperGrok** | $30/month | Grok 4 via grok.com, highest quotas |
| **SuperGrok Heavy** | $300/month | Grok 4 Heavy (multi-agent, max compute) |

These are conversational-only and do **not** provide API access. The API is a separate product with its own billing.

---

## Additional Tips

- **OpenAI SDK compatibility** -- You can use the OpenAI Python/JS SDK by changing the base URL to `https://api.x.ai/v1`. This makes migration trivial if you are already using OpenAI
- **Spending limits** -- Configure your invoiced spending limit at console.x.ai. The default is $0, which stops API access when credits run out (no surprise bills)
- **Grok 4 is reasoning-only** -- Unlike other models, Grok 4 always uses chain-of-thought reasoning. There is no non-reasoning mode, and the `reasoning_effort` parameter is not supported (it will return an error)
- **Image input** -- Vision-capable models accept images up to 20 MB in jpg/jpeg/png format with no limit on quantity
- **Knowledge cutoff** -- Grok 3 and Grok 4 have a knowledge cutoff of November 2024, but Live Search and Web Search tools can fetch real-time information
- **SpaceX acquisition** -- As of February 2026, SpaceX has acquired xAI. This may affect future pricing and platform direction
- **Multiple teams strategy** -- Create separate teams for sensitive vs. non-sensitive projects. Only opt into data sharing on teams handling non-confidential data
- **Credit card not required** -- For the $25 signup credit and initial exploration, no payment method is needed. You only need to add a payment method for prepaid credits or invoiced billing beyond free credits

---

**Sources:**
- [xAI API Overview](https://docs.x.ai/docs/overview)
- [xAI Models and Pricing](https://docs.x.ai/docs/models)
- [xAI API Console](https://console.x.ai)
- [xAI API Page](https://x.ai/api)
- [xAI Hitchhiker's Guide](https://docs.x.ai/docs/tutorial)
- [xAI Grok API Pricing Guide 2026 (aifreeapi.com)](https://www.aifreeapi.com/en/posts/xai-grok-api-pricing)
- [Grok API Pricing Calculator (costgoat.com)](https://costgoat.com/pricing/grok-api)
- [xAI Data Sharing Program (cloudcredits.io)](https://cloudcredits.io/providers/xai/programs/data-sharing-program)
- [Grok AI Supported Countries (prexility.com)](https://prexility.com/grok-ai-supported-countries-and-regions/)
- [xAI Pricing Explained (eesel.ai)](https://www.eesel.ai/blog/xai-pricing)
- [Grok Free Plans and Subscriptions (datastudios.org)](https://www.datastudios.org/post/grok-ai-free-plans-trials-and-subscriptions-structure-pricing-and-model-access-in-2025)

## Submission 73: Vercel AI Accelerator - $4M+ Combined Credits

- Category: Startup Programs
- Source Link: https://vercel.com/ai-accelerator
- Amount: $4M+ combined credits (Vercel + AWS + 20+ AI partners per cohort)
- Difficulty: 5/5 (Very Hard)
- Expires At: N/A

### Description

The Vercel AI Accelerator is a **six-week program for 40 selected AI startups** offering over **$4M in combined credits** from Vercel, v0, AWS, and 20+ AI platform partners. Every accepted team gets thousands of dollars in credits from each partner, with three Demo Day winners earning **$100K+ each** in additional prizes. The first-place winner also receives an **investment from Vercel Ventures**. The program runs cohorts periodically (Cohort 1 in 2023, Cohort 3 in mid-2025) and is highly competitive -- roughly **2.5% acceptance rate** (40 out of 1,500+ applicants in Cohort 1). No credit card required; this is a selective application-based program, not a self-serve offer.

---

## How to Apply (Step-by-Step)

1. Go to [vercel.com/ai-accelerator](https://vercel.com/ai-accelerator)
2. Click the **"Apply"** button, which opens a **Typeform** application
3. Fill out the application -- expect questions about your AI product, team, current traction, tech stack, and how you plan to use the credits
4. Submit before the deadline (e.g. May 17, 2025 for Cohort 3)
5. Wait for selection -- Vercel typically announces the 40 accepted teams about two weeks after the deadline
6. If accepted, you receive onboarding instructions and credit codes from each partner

**Important:**
- Applications are only open during specific windows (typically 2-4 weeks before each cohort)
- No fee to apply or participate
- No credit card or payment information required
- The program is **virtual-first** with an optional in-person meetup in San Francisco
- Acceptance is not guaranteed -- prepare a strong application showcasing your AI product and traction

---

## Eligibility & Selection

- **Early-stage startups** (typically pre-seed through Series A) building AI-powered products
- Open **globally** -- no region restrictions
- **Existing Vercel Enterprise customers are not eligible**
- Individuals and teams can both apply (past winners include solo builders)
- Preference may be given to teams already building on Vercel/Next.js, but it is not a strict requirement
- Having a working product or MVP strengthens your application significantly
- Past cohorts accepted a mix of startups at various stages, from pre-launch MVPs to products with hundreds of thousands of users

---

## Credits Breakdown by Partner (Cohort 3, 2025)

The total pool exceeds **$4M** distributed across all 40 participants. Below are the confirmed partner allocations:

| Partner | Total Pool | Per Participant | Demo Day Winner Bonus |
|---|---|---|---|
| **AWS** (Activate credits) | $1,000,000 | $25,000 | -- |
| **Vercel** (cloud credits) | $1,000,000+ | $1,000+ | $2,000 |
| **Anthropic** (API credits) | $600,000 | $15,000 | 3 months Claude for Teams |
| **Daytona** | $625,000 | $10,000 | $75,000 |
| **Perplexity** (API credits) | $475,000 | $10,000 | $10,000 + 70% first-year discount |
| **Modal** (compute) | $350,000 | $5,000 | $50,000 |
| **v0** (platform credits) | $216,000 | $40+ | $200 |
| **Cartesia** | $181,000 | $4,000 | $4,000 |
| **ElevenLabs** | $170,000 | $4,000 | $4,000 |
| **OpenAI** (API credits) | $100,000 | varies | varies |

### Additional Partners (credits vary)

Braintrust, Cursor, Fal, Groq, Hugging Face, Hyperbolic ($50/participant, $1,000 winner), Julius AI, Langbase + Chai, Mistral AI, Patronus, Quotient AI, Roboflow ($150K total pool), Sentry, Sieve, and Vapi.

**Note:** The exact per-participant amount varies by partner and cohort. Partner lists and credit amounts change between cohorts as new sponsors join and others rotate out. The figures above reflect the Cohort 3 (2025) offering.

---

## Program Structure

| Week | Activities |
|---|---|
| **Week 1** | Onboarding, credit activation, kickoff meetup (optional, San Francisco) |
| **Weeks 1-5** | Weekly fireside chats with AI founders, VCs, and engineers |
| **Weeks 1-5** | Office hours with mentors and industry leaders |
| **Weeks 1-6** | Building in private Discord community with other participants |
| **Week 6** | Demo Day preparation and pitch practice |
| **Final Day** | Demo Day in San Francisco (28 of 40 teams pitch, 3 minutes each) |

### Past Fireside Chat Speakers & Judges

- **Nat Friedman** (Investor, former GitHub CEO)
- **Guillermo Rauch** (Vercel CEO)
- **Lin Qiao** (CEO of Fireworks.ai)
- **Shyamal Anadkat** (OpenAI GTM)
- 15+ additional AI experts and VCs across weekly sessions

---

## Demo Day & Prizes

At the end of the six weeks, **28 of the 40 teams** are selected to pitch at Demo Day in San Francisco. Each team gets **3 minutes** to present to a panel of judges and an audience of **24+ VCs**.

Three winners are selected across categories:

| Award | Prize |
|---|---|
| **Best Overall** | $100K+ in additional credits + **Vercel Ventures investment** |
| **Best Startup Team** | $100K+ in additional credits |
| **Best Individual Team** | $100K+ in additional credits |

Past Cohort 1 winners each received **$120K in credits**. Cohort 3 winner bonuses from individual partners are listed in the credits table above.

---

## What Happens After the Program

- **Credits expire** based on each partner's terms -- durations vary (some 3 months, some 12 months, some usage-based)
- You retain access to the **alumni community** and network
- Vercel services revert to regular pricing after credits are consumed
- There is **no equity taken** by Vercel for participating (the Vercel Ventures investment for the first-place winner is a separate, voluntary deal)
- No obligation to continue using Vercel after the program ends

---

## Past Cohort Results

Cohort 1 (2023) notable outcomes:

| Winner | Achievement During Program |
|---|---|
| **Memorang** (Best Overall) | Added $1M+ ARR |
| **Jenni** (Best Startup Team) | Added $500K ARR, hundreds of thousands of users |
| **LLM Report** (Best Individual) | Shipped open-source LLM observability platform |

Some participants added hundreds of thousands of users; others built and launched an entire MVP within the six weeks.

---

## Tips & Gotchas

- **Highly competitive** -- less than 3% acceptance rate. Invest time in your application; a working demo or strong traction helps
- **No equity taken** for participation -- only the first-place winner negotiates a Vercel Ventures investment separately
- **Not a traditional accelerator** -- there is no cash investment for all participants, only credits and mentorship
- **Virtual-first** -- you do not need to relocate. The San Francisco meetup and Demo Day are optional (though Demo Day attendance is strongly recommended)
- **Credit durations vary by partner** -- some expire in weeks, others in months. Activate and plan usage early
- **You do not need to use Vercel** for your stack to apply, but using Next.js or Vercel may strengthen your application
- **Alternative:** If you miss the Accelerator window, check [Vercel for Startups](https://vercel.com/startups) for smaller credit packages ($2,400+) available through VC partner programs year-round
- **Cohorts are periodic** -- if applications are closed, follow [@vercel](https://x.com/vercel) on X for announcements about the next cohort

---

**Sources:**
- [Vercel AI Accelerator](https://vercel.com/ai-accelerator)
- [Join the Vercel AI Accelerator (Cohort 3 Blog Post)](https://vercel.com/blog/join-the-vercel-ai-accelerator)
- [Introducing Vercel's AI Accelerator (Original Announcement)](https://vercel.com/blog/vercel-ai-accelerator)
- [Vercel AI Accelerator Demo Day (Cohort 1 Results)](https://vercel.com/blog/vercel-ai-accelerator-demo-day)
- [My Experience Applying for the Vercel AI Accelerator (Indie Hackers)](https://www.indiehackers.com/product/prayetic/my-experience-applying-for-the-vercel-ai-accelerator--NaI49AxFza4kqf8YXTQ)
- [Vercel for Startups](https://vercel.com/startups)

## Submission 72: Together AI Startup Accelerator - Up to $50,000

- Category: Startup Programs
- Source Link: https://www.together.ai/startup-accelerator
- Amount: Up to $50,000 in platform credits (tiered: $15K / $30K / $50K based on funding stage)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Together AI offers a **selection-based accelerator program** providing up to **$50,000 in free platform credits** for AI-native startups. Credits cover serverless inference, fine-tuning, dedicated endpoints, and instant GPU clusters. The program is structured into **three funding-based tiers** (Build, Grow, Scale) and includes dedicated AI engineering support, go-to-market amplification, exclusive founder events, and VC network access. This is not a self-serve signup — you apply and Together AI's team reviews your application.

---

## Program Tiers

The accelerator is split into three tiers based on how much funding your startup has raised:

| Tier | Funding Stage | Platform Credits | Engineering Hours |
|---|---|---|---|
| **Build** | Up to $5M raised | Up to **$15,000** | 3 hours |
| **Grow** | $5M–$10M raised | Up to **$30,000** | 6 hours |
| **Scale** | Over $10M raised | Up to **$50,000** | 10 hours |

**Important:** Credits **exclude Reserved GPU Clusters**. They can be used for serverless inference, fine-tuning, dedicated endpoints, and instant clusters.

---

## What You Get

### Platform Credits
Free credits to use across Together AI's core products:
- **Serverless Inference** — pay-per-token access to 200+ open-source models (Llama, DeepSeek, Qwen, Mistral, and more) via an OpenAI-compatible API
- **Fine-Tuning** — supervised fine-tuning and DPO (Direct Preference Optimization) with LoRA on models like DeepSeek-R1, DeepSeek-V3, Llama, and others. You own the resulting model 100%
- **Dedicated Endpoints** — single-tenant GPU instances (H100, H200) for consistent performance without shared compute. Per-minute billing, starting around $2.10–$4.99/hour depending on hardware
- **Instant GPU Clusters** — on-demand clusters with Kubernetes or Slurm, InfiniBand networking, for large-scale training or inference

### Forward-Deployed Engineering
Dedicated time with Together AI engineers who help you:
- Optimize your model architecture and inference stack
- Address performance bottlenecks
- Set up fine-tuning pipelines and evaluation workflows
- Implement best practices for benchmarking

### Go-to-Market (GTM) Support
- Marketing amplification and co-selling opportunities to Together AI's **1M+ developer network**
- Joint case studies, blog features, and event participation

### Community and VC Access
- Exclusive founder events with AI researchers (including notable figures like Tri Dao and Percy Liang)
- Introductions to Together AI's network of venture capital partners
- Peer networking with other accelerator cohort members

---

## Available Models (via Platform Credits)

Together AI provides access to **200+ open-source models**. Key models include:

| Category | Notable Models |
|---|---|
| **Large Language Models** | Llama 4 Scout/Maverick, Llama 3.3 70B, DeepSeek-R1, DeepSeek-V3, Qwen 2.5 (various sizes), Mistral, Kimi K2 |
| **Code Models** | Qwen 2.5 Coder, DeepSeek Coder |
| **Vision Models** | Llama 3.2 Vision, Qwen-VL |
| **Image Generation** | FLUX (various versions) |
| **Embeddings & Reranking** | Various embedding and rerank models |

All endpoints are **OpenAI-compatible**, so you can swap in Together AI with minimal code changes if you are already using OpenAI's API format.

---

## Eligibility & Selection Criteria

- **AI-native startups** building applications (not infrastructure or research-only projects)
- Selection is based on **funding stage**, customer adoption, technical requirements, and alignment with open-source AI development
- The application asks for: company fundamentals, funding history, customer metrics, GPU requirements, deployment timeline, current cloud providers, model architectures in use, monthly GPU spending, and specific use cases
- No specific geographic restrictions mentioned — the program appears to be **global**
- No credit card or payment required to apply

---

## How to Apply

1. Go to [together.ai/startup-accelerator](https://www.together.ai/startup-accelerator)
2. Fill out the application form with your startup details (company info, funding stage, tech stack, use case)
3. Together AI's team reviews your application — **no public timeline** is given for review turnaround
4. If accepted, you are placed into the appropriate tier (Build, Grow, or Scale) based on your funding level
5. Credits are provisioned to your Together AI account and engineering hours are scheduled

**Note:** This is a rolling application — there are no fixed cohort deadlines mentioned. Startups like Corridor.dev and PlayerZero are already part of the program.

---

## Credit Exclusions & Restrictions

- Credits **do not cover Reserved GPU Clusters** (only Instant Clusters are included)
- Together AI's general platform requires a **minimum $5 credit purchase** to unlock full API access (Build Tier 1). It is unclear whether accelerator credits bypass this requirement — ask during the application process
- No free trial exists on the general platform outside of this program
- Credits are likely time-limited (specific expiration terms are set during onboarding)

---

## Approximate Credit Value

To give you a sense of what the credits are worth in practice:

| Use Case | Approximate Cost | $15K Credits Gets You |
|---|---|---|
| **Llama 3.3 70B inference** | ~$0.88/M tokens (input+output) | ~17 billion tokens |
| **DeepSeek-R1 inference** | ~$3.50/M tokens (input) | ~4.3 billion input tokens |
| **Fine-tuning (Llama 70B)** | ~$2.90/M tokens | ~5.2 billion training tokens |
| **Dedicated Endpoint (H100)** | ~$2.99/hour | ~5,000 GPU-hours |
| **Image Generation (FLUX)** | ~$0.03/image | ~500,000 images |

These are rough estimates — actual costs depend on model, batch size, and usage pattern.

---

## Related Programs

- **Together AI Research Credits Program** — small grants for students conducting independent research projects. Apply at [together.ai/forms/research-credits-program-request](https://www.together.ai/forms/research-credits-program-request)
- **Together Fund AI Studio** — a separate VC-backed program (together.fund, not together.ai) offering up to $600K in credits across multiple cloud providers. 12-week cohort program with a different application process
- **Promotional Credits** — Together AI occasionally runs promotional campaigns (e.g., $100 free credits on signup), but these are not always active

---

## Additional Tips

- **Not a traditional accelerator** — Together AI does not take equity or require you to join a cohort in person. It is essentially a credits + support program branded as an accelerator
- **Open-source focus** — the platform specializes in open-source models. If you need proprietary models (GPT-4, Claude, etc.), this program does not cover them
- **Compare with alternatives** — AWS Generative AI Accelerator, Google for Startups Cloud Program, and Microsoft for Startups also offer substantial credits. Together AI's advantage is competitive pricing on open-source model inference (up to 11x cheaper than GPT-4o for equivalent workloads)
- **Engineering hours are valuable** — the forward-deployed engineering time is arguably the most unique benefit. Use it to optimize your most critical inference or fine-tuning workloads
- **SOC 2 and HIPAA compliant** — if your startup handles sensitive data, Together AI supports compliant deployments including private VPC options

---

**Sources:**
- [Together AI Startup Accelerator](https://www.together.ai/startup-accelerator)
- [Announcing the Together AI Startup Accelerator (Blog)](https://www.together.ai/blog/announcing-together-ai-startup-accelerator)
- [Together AI Pricing](https://www.together.ai/pricing)
- [Together AI Products](https://www.together.ai/products)
- [Together AI Billing & Usage Limits](https://docs.together.ai/docs/billing)
- [Together AI Research Credits Program](https://www.together.ai/forms/research-credits-program-request)

## Submission 71: Snowflake for Startups - Credits Program

- Category: Startup Programs
- Source Link: https://www.snowflake.com/en/why-snowflake/startup-program
- Amount: 10% of annual commitment as free credits (e.g., $10K on a $100K contract)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Snowflake for Startups gives early-stage companies **free Snowflake credits equal to 10% of their annual commitment** (minimum $10K ACV), plus dedicated technical support, go-to-market resources, co-selling opportunities, and access to a global VC network. The program evolved from the earlier Powered by Snowflake initiative and was relaunched in September 2025 with expanded benefits, including dedicated AI inference capacity through Cortex AI and a six-month Startup Accelerator backed by $200M in investment capital.

---

## Eligibility

- **Stage:** Pre-Seed through Series C startups founded within the last 10 years
- **Customer status:** New Snowflake customers. Existing month-to-month customers can qualify by switching to an annual contract
- **Minimum commitment:** $10,000 annual contract value (ACV)
- **Product focus:** Must be building a data or AI application on Snowflake's AI Data Cloud
- **Region:** Global — no geographic restrictions mentioned

There is no explicit requirement for VC backing or a specific revenue threshold. The program is oriented toward startups that plan to use Snowflake as a core part of their product architecture, not just as an internal analytics tool.

---

## Credit Structure

The credit math is straightforward:

| Annual Commitment (ACV) | Free Credits (10%) |
|---|---|
| $10,000 | $1,000 |
| $50,000 | $5,000 |
| $100,000 | $10,000 |
| $250,000 | $25,000 |

Credits are added on top of whatever discounts the startup program already provides. They are tied to the annual contract period — when the contract renews, the credit structure resets.

**Important:** This is not a free trial. You must sign an annual contract and commit real spend. The 10% credits are a bonus on top of that commitment, not a standalone grant.

---

## Registration (Step-by-Step)

1. Go to [snowflake.com/startup-program](https://www.snowflake.com/en/why-snowflake/startup-program/)
2. Click the application or contact form to express interest
3. Provide details about your startup: stage, product, how you plan to use Snowflake
4. Snowflake's startup team reviews your application and reaches out
5. If approved, you negotiate an annual contract (minimum $10K ACV)
6. Free credits (10% of ACV) are added to your contract upon signing

**Alternative path — through a VC partner:**
- If your startup is backed by one of Snowflake's partner VCs (Coatue, Greylock, Madrona, Capital One Ventures, Bain Capital Ventures, etc.), you may get a faster or more favorable onboarding through their portfolio program
- For example, Aurelia Ventures portfolio companies access credits through their Founders Hub member portal

**Caveats:**
- This is not an instant self-service signup — expect a sales conversation
- Approval timelines are not published; plan for days to weeks depending on your stage and use case
- Month-to-month accounts must convert to annual contracts to qualify

---

## What You Get Beyond Credits

### Technical Support
- Dedicated **account manager** and **sales engineer**
- Access to expert **application architects** who help you design your Snowflake-based product
- AI/ML technical support for Cortex AI workloads

### Go-to-Market Support
- **Co-selling opportunities** through Snowflake Accelerate (Snowflake's marketplace and partner channel)
- Self-service **co-marketing resources** designed for early-stage companies
- Inclusion in **startup showcases**, user groups, and events
- Visibility within Snowflake's global ecosystem of 12,600+ customers

### Community
- Access to Snowflake's **Slack community** and user groups
- Invitations to startup-specific events and meetups

---

## Cortex AI — Available Models & Services

Startups in the program get access to Snowflake Cortex AI, which provides managed inference for leading frontier models — all running within Snowflake's security perimeter so your data never leaves the platform.

### Available LLM Providers & Models

| Provider | Models |
|---|---|
| **Anthropic** | Claude 4 Opus, Claude 4 Sonnet, Claude 3.7 Sonnet, Claude 3.5 Sonnet |
| **OpenAI** | GPT-4.1, o4-mini, GPT-5.2 (private preview) |
| **Meta** | Llama 3 (8B, 70B), Llama 3.1 (8B, 70B, 405B), Llama 3.3 70B, Llama 4 Maverick, Llama 4 Scout |
| **Mistral AI** | Mistral Large, Mistral Large 2, Mistral 7B, Mixtral 8x7B |
| **DeepSeek** | DeepSeek-R1 |
| **Google** | Gemini models (coming natively to Cortex) |
| **Snowflake** | Arctic, Snowflake-Llama-3.1-405B, Snowflake-Llama-3.3-70B |

### Cortex AI Services

| Service | Description |
|---|---|
| **Cortex Search** | Semantic search over your Snowflake data |
| **Cortex Analyst** | Natural language querying of structured data |
| **Document AI** | Extract structured data from documents |
| **Cortex Code** | AI coding agent for data-centric workflows (GA) |
| **Snowflake Intelligence** | Conversational agentic application for enterprise data |
| **LLM Functions** | SQL-callable AI (COMPLETE, SUMMARIZE, TRANSLATE, SENTIMENT, etc.) |
| **Cortex REST API** | OpenAI-compatible REST API for programmatic access |

All models run inside Snowflake's compute — your data stays in your account and is not sent to external APIs.

---

## Startup Accelerator (Separate Track)

The **Snowflake Startup Accelerator** is a more selective, six-month program for early-stage companies. It is distinct from the general Snowflake for Startups credits program.

**What you get:**
- Cloud credits (amount varies)
- Go-to-market support and co-marketing
- Technical guidance from Snowflake engineers
- AWS credits (up to $1M in free Snowflake credits on AWS over four years, through AWS partnership)
- Dedicated AI inference capacity
- Potential investment from Snowflake Ventures ($200M fund)
- Exposure to partner VCs

**Scale:**
- Winter 2026 cohort: 49 startups accepted from 500+ applications
- Applications increased 300%+ year-over-year
- 29 products launched through the program
- Notable alumni include Coalesce, LandingAI (Andrew Ng), and Twelve Labs

**How to apply:** Visit [snowflake.com/startup-accelerator](https://www.snowflake.com/en/why-snowflake/startup-program/startup-accelerator/)

---

## Startup Challenge 2026 (Competition)

Separate from the credits program, Snowflake runs an annual **Startup Challenge** in collaboration with the NYSE.

- **Prize:** Up to $500,000 in potential investment for the winner; up to $250,000 each for two other finalists
- **Eligibility:** Early-stage startups that have raised no more than $5M in cash funding, no priced equity rounds (Series A+), must be building on Snowflake
- **Requirement:** Submit a 5-minute demo video
- **Details:** [snowflake.com/startup-challenge](https://www.snowflake.com/en/startup-challenge/)

---

## VC Network

The program connects startups with a network of 12+ venture capital firms:

- **Coatue**
- **Greylock Partners**
- **Madrona Ventures**
- **Capital One Ventures**
- **Bain Capital Ventures**
- **Blackstone Innovations Investments**
- **Bessemer Venture Partners**
- **General Catalyst**
- **Hetz Ventures**
- **Mayfield**
- **NewBuild Venture Capital**
- **NTTVC**
- **Virtue**

Snowflake Ventures itself has 65+ portfolio companies and exits, including 15+ early-stage investments made through the startup programs.

---

## Silicon Valley AI Hub

Snowflake opened the **SVAI Hub** in Menlo Park, California — a 28,000-square-foot facility offering:
- Coworking spaces for startups in the program
- AI-focused educational events
- Collaboration space for startups, investors, and Snowflake engineers

Access is available to startups participating in the accelerator or broader startup program.

---

## Additional Tips

- **This is not a free tier** — unlike AWS, GCP, or Azure free trials, Snowflake for Startups requires a real annual contract. The 10% credit bonus is a discount mechanism, not a standalone grant
- **Snowflake pricing is consumption-based** — you pay for compute (credits) and storage separately. Your $10K minimum commitment buys compute credits that deplete as you run queries and workloads
- **No credit card trial** — Snowflake does offer a separate [30-day free trial](https://signup.snowflake.com/) with $400 in credits for anyone (no startup program needed), but it is unrelated to this program
- **Multi-cloud** — Snowflake runs on AWS, Azure, and GCP. Your credits work regardless of which cloud provider hosts your Snowflake account
- **Region availability for AI models** — not all Cortex AI models are available in every Snowflake region. Cross-region inference is available for some models but check documentation for specifics
- **Lock-in consideration** — building your product on Snowflake's platform creates significant switching costs. Evaluate whether Snowflake is the right long-term data platform before committing

---

**Sources:**
- [Snowflake for Startups](https://www.snowflake.com/en/why-snowflake/startup-program/)
- [Snowflake Startup Accelerator](https://www.snowflake.com/en/why-snowflake/startup-program/startup-accelerator/)
- [Snowflake Launches New Program for Startups — SiliconANGLE](https://siliconangle.com/2025/09/24/snowflake-launches-new-snowflake-startups-program/)
- [Snowflake Grows Startup Accelerator with $200M — TechCrunch](https://techcrunch.com/2025/02/27/snowflake-grows-startup-accelerator-with-200m-in-new-capital/)
- [Snowflake Credits via Aurelia Ventures](https://aureliaventures.com/partners/snowflake)
- [Snowflake Cortex AI](https://www.snowflake.com/en/product/features/cortex/)
- [Snowflake Cortex AI LLM Functions Documentation](https://docs.snowflake.com/en/user-guide/snowflake-cortex/aisql)
- [Snowflake Startup Challenge 2026](https://www.snowflake.com/en/startup-challenge/)
- [OpenAI GPT-5.2 on Snowflake Cortex AI](https://www.snowflake.com/en/blog/announcing-openai-snowflake-cortex-ai/)
- [Meet the Startups Building AI on Snowflake](https://www.snowflake.com/en/blog/startups-building-ai-snowflake/)

## Submission 70: SambaNova Cloud - Free API Access

- Category: AI API Free Tiers
- Source Link: https://cloud.sambanova.ai/
- Amount: $5 free credit (3-month expiry) + ongoing rate-limited Free Tier (40 RPD, 200K TPD)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

SambaNova Cloud offers **free API access** to top open-source models including DeepSeek R1, DeepSeek V3, Llama 4 Maverick, Qwen 3, and more — powered by SambaNova's custom RDU chips that deliver inference speeds up to **10x faster than GPUs**. No credit card is required to start. New accounts receive **$5 in free credits** (valid for 3 months), which translates to over 30 million tokens on Llama 8B. After credits are exhausted, a rate-limited Free Tier remains available (10-40 RPM depending on model, 40 RPD). The API is **OpenAI-compatible**, so you can swap in SambaNova as a drop-in replacement in most codebases.

---

## Registration (Step-by-Step)

1. Go to [cloud.sambanova.ai](https://cloud.sambanova.ai)
2. Click **"Get Started"** or **"Sign Up"**
3. Create an account (email signup or SSO)
4. Once logged in, navigate to the **API Keys** page at [cloud.sambanova.ai/apis](https://cloud.sambanova.ai/apis)
5. Click **"Create New Key"** — save it immediately, as keys cannot be viewed again after creation
6. Set the environment variable `SAMBANOVA_API_KEY` or pass it directly in your API calls
7. Use the base URL `https://api.sambanova.ai/v1` — compatible with OpenAI client libraries

**Important:**
- No credit card or payment method required for the Free Tier
- You can create up to **25 API keys** per account
- The $5 free credit is automatically applied to your account and **expires after 3 months**
- After credits expire, you remain on the Free Tier with rate-limited access (no payment needed)
- Linking a payment method upgrades you to the Developer Tier with higher rate limits

---

## Available Models

### Production Models (stable, recommended for apps)

| Model | Type | Notes |
|---|---|---|
| **DeepSeek-R1-0528** | Reasoning | 671B MoE, advanced reasoning/math/code |
| **DeepSeek-R1-Distill-Llama-70B** | Reasoning | Distilled reasoning model |
| **DeepSeek-V3-0324** | Chat | 671B MoE, general-purpose |
| **DeepSeek-V3.1** | Chat | Latest DeepSeek generation |
| **DeepSeek-V3.1-Terminus** | Chat | Enhanced V3.1 variant |
| **DeepSeek-V3.2** | Chat | Newest DeepSeek model |
| **Meta-Llama-3.1-8B-Instruct** | Chat | Fast, lightweight |
| **Meta-Llama-3.3-70B-Instruct** | Chat | Strong general-purpose |

### Preview Models (experimental, may be removed)

| Model | Type | Notes |
|---|---|---|
| **Llama-4-Maverick-17B-128E-Instruct** | Chat/Vision | 17B active params, 128 experts MoE |
| **openai/gpt-oss-120b** | Chat | Open-source 120B MoE from OpenAI |
| **Qwen3-32B** | Chat | Alibaba's Qwen 3 series |
| **Qwen3-235B** | Chat | Large Qwen 3 model |
| **E5-Mistral-7B-Instruct** | Embeddings | Text embeddings model |
| **Whisper-Large-v3** | Audio | Speech-to-text transcription |
| **Llama-3.3-Swallow-70B-Instruct** | Chat | Japanese-focused Llama variant |
| **ALLaM-7B-Instruct** | Chat | Arabic language model |

**Note:** Preview models have limited capacity and may be removed at short notice. Do not rely on them for production workloads.

---

## Rate Limits

### Free Tier (no payment method linked)

| Model | RPM | RPD | TPD |
|---|---|---|---|
| Meta-Llama-3.1-8B-Instruct | 40 | 40 | 200,000 |
| Meta-Llama-3.3-70B-Instruct | 40 | 40 | 200,000 |
| DeepSeek-R1 | 20 | 40 | 200,000 |
| DeepSeek-R1-Distill-Llama-70B | 40 | 40 | 200,000 |
| DeepSeek-V3-0324 | 20 | 40 | 200,000 |
| DeepSeek-V3.1 | 20 | 40 | 200,000 |
| Llama-4-Maverick-17B-128E | 20 | 40 | 200,000 |
| gpt-oss-120b | 20 | 40 | 200,000 |
| Qwen3-32B | 10 | 40 | 200,000 |
| Whisper-Large-v3 | 40 | 40 | 200,000 |
| E5-Mistral-7B-Instruct | 20 | 40 | 200,000 |

*RPM = Requests per Minute, RPD = Requests per Day, TPD = Tokens per Day*

### Developer Tier (payment method linked, $5 free credit included)

| Model | RPM | RPD |
|---|---|---|
| Meta-Llama-3.1-8B-Instruct | 1,440 | 288,000 |
| Meta-Llama-3.3-70B-Instruct | 240 | 48,000 |
| DeepSeek-R1 | 60 | 12,000 |
| DeepSeek-R1-Distill-Llama-70B | 240 | 48,000 |
| DeepSeek-V3-0324 | 60 | 12,000 |
| DeepSeek-V3.1 | 60 | 12,000 |
| Llama-4-Maverick-17B-128E | 60 | 12,000 |
| gpt-oss-120b | 60 | 12,000 |
| Qwen3-32B | 30 | 6,000 |
| Whisper-Large-v3 | 450 | 90,000 |
| E5-Mistral-7B-Instruct | 60 | 12,000 |

The Developer Tier provides **dramatically higher limits** — for example, Llama 8B goes from 40 to 1,440 RPM (36x increase). The key difference: the Free Tier caps you at **40 requests per day** and **200,000 tokens per day** across all models.

---

## Pricing (Developer Tier, per million tokens)

| Model | Input | Output |
|---|---|---|
| Meta-Llama-3.1-8B-Instruct | $0.01 | $0.02 |
| Meta-Llama-3.3-70B-Instruct | $0.06 | $0.12 |
| DeepSeek-V3.1 (cb) | $0.015 | $0.075 |
| DeepSeek-V3.1-Terminus | $0.30 | $0.45 |
| DeepSeek-V3.2 | $0.30 | $0.45 |
| DeepSeek-R1-0528 | $0.50 | $0.70 |
| DeepSeek-R1-Distill-Llama-70B | $0.07 | $0.14 |
| Llama-4-Maverick-17B-128E | $0.063 | $0.18 |
| openai/gpt-oss-120b | $0.022 | $0.059 |
| Qwen3-32B | $0.04 | $0.08 |
| Qwen3-235B | $0.04 | $0.08 |
| E5-Mistral-7B-Instruct | $0.013 | Free |
| Whisper-Large-v3 | $10/hour | N/A |

With the $5 free credit, you can run roughly **30 million tokens** on Llama 8B, or about **70,000 tokens** on DeepSeek-R1-0528.

---

## Free Tier vs. Developer Tier

| Feature | Free Tier | Developer Tier |
|---|---|---|
| **Payment method** | Not required | Required (credit card) |
| **Free credit** | $5 (expires in 3 months) | $5 (expires in 3 months) |
| **RPM** | 10-40 depending on model | 30-1,440 depending on model |
| **RPD** | 40 across all models | 6,000-288,000 depending on model |
| **TPD** | 200,000 across all models | No stated cap |
| **After credits expire** | Rate-limited access continues | Pay-as-you-go |
| **Production use** | Not practical (40 RPD cap) | Yes |

---

## The Catch: Free Tier Limitations

The Free Tier's biggest constraint is the **40 requests per day** limit across all models, combined with a **200,000 tokens per day** cap. This means:

- You can do about **40 short conversations per day** for testing and prototyping
- The RPM limits (10-40) are less of an issue than the daily caps
- This is enough for **experimentation and evaluation**, but not for building anything real
- SambaNova has stated they have **no plans to maintain the free tier** long-term — it may be fully replaced by the credit-based Developer Tier

If you want meaningful free usage, the **$5 credit on the Developer Tier** (no charge until credits run out) is the better path. Just be aware that linking a payment method means you can be billed once credits are exhausted.

---

## OpenAI-Compatible API

SambaNova's API is fully **OpenAI-compatible**, which means you can use standard OpenAI client libraries:

```python
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_SAMBANOVA_API_KEY",
    base_url="https://api.sambanova.ai/v1"
)

response = client.chat.completions.create(
    model="DeepSeek-R1",
    messages=[{"role": "user", "content": "Hello!"}]
)
```

This makes it trivial to switch between SambaNova, OpenAI, Together AI, and other compatible providers.

---

## Earning Additional Credits

- **Community participation** — the top 10 most engaged members on the [SambaNova Developer Community](https://community.sambanova.ai) each month receive **$10 in additional credits**
- **Community membership bonus** — existing community members received an extra $5 credit when the Developer Tier launched
- Engagement is tracked via a community leaderboard (answering questions, sharing projects, providing feedback)

---

## Additional Tips

- **Speed advantage** — SambaNova's RDU hardware delivers extremely fast inference. DeepSeek-R1 671B runs at ~200 tokens/sec vs. ~19 tokens/sec average on GPU providers. If speed matters to you, this is a strong differentiator
- **No auto-billing on Free Tier** — without a payment method linked, you will never be charged. The Free Tier simply rate-limits you after credits expire
- **Developer Tier auto-billing** — once you link a payment method, you will be billed for usage beyond the $5 free credit. Set spending alerts if available
- **Preview models are unstable** — models marked "Preview" (including Llama 4 Maverick, gpt-oss-120b, Qwen3) may be removed without notice. Don't build production features on them
- **AWS Marketplace** — SambaCloud is also available through AWS Marketplace if you prefer consolidated billing
- **Rate limit headers** — every API response includes `X-RateLimit-Remaining` and related headers so you can programmatically track your usage
- **Free tier future uncertain** — SambaNova has indicated they may fully deprecate the Free Tier in favor of the credit-based Developer Tier. Sign up while it is still available

---

**Sources:**
- [SambaNova Cloud Plans](https://cloud.sambanova.ai/plans)
- [SambaNova Cloud Pricing](https://cloud.sambanova.ai/plans/pricing)
- [SambaNova Rate Limits Documentation](https://docs.sambanova.ai/docs/en/models/rate-limits)
- [SambaNova Developer Tier Blog Post](https://sambanova.ai/blog/sambanova-cloud-developer-tier-is-live)
- [SambaNova API Keys & URLs](https://docs.sambanova.ai/docs/en/get-started/api-keys-urls)
- [SambaNova Developer Community - Free Tier Discussion](https://community.sambanova.ai/t/is-free-tier-going-away/847)
- [SambaNova Supported Models](https://docs.sambanova.ai/cloud/docs/get-started/supported-models)
- [Free LLM API Resources (GitHub)](https://github.com/cheahjs/free-llm-api-resources)

## Submission 69: SambaNova Developer Tier - $5 Free Credits

- Category: AI API Free Tiers
- Source Link: https://cloud.sambanova.ai/plans
- Amount: $5 in free credits (~30M+ tokens on Llama 8B, 500M+ tokens on cheapest models)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

SambaNova Cloud gives every new developer **$5 in free credits** (roughly 30 million tokens on Llama 3.1 8B) to explore their ultra-fast inference platform. SambaNova runs open-source models like DeepSeek, Llama, and Qwen on custom RDU hardware, delivering inference speeds up to 10x faster than GPU-based providers. The Developer Tier offers **50% higher rate limits** on 70B/405B models and **doubled rate limits** on Llama 8B compared to the Free Tier. After credits run out, it switches to pay-as-you-go billing. No approval process, no waitlist.

---

## Registration (Step-by-Step)

1. Go to [cloud.sambanova.ai](https://cloud.sambanova.ai/)
2. Click **"Get Started"** or **"Sign Up"**
3. Create an account using your **email address** (or sign in if you already have one)
4. Verify your email if prompted
5. You land on the dashboard with **$5 in free credits** already applied
6. Go to [API Keys](https://cloud.sambanova.ai/apis) and click **"Create new key"** — save it immediately, as it cannot be viewed again
7. To unlock **Developer Tier rate limits**, go to [Plans > Billing](https://cloud.sambanova.ai/plans/billing) and **add a payment method** (credit card via Stripe)
8. Start calling models using the **OpenAI-compatible API** endpoint

**Important:**
- Without a payment method, you stay on the **Free Tier** with lower rate limits and a 200,000 tokens-per-day cap
- Adding a payment method upgrades you to the **Developer Tier** with significantly higher rate limits
- You can generate and use up to **25 API keys** per account
- The API is **OpenAI-compatible** — just swap the base URL and API key in any OpenAI SDK

---

## Free Tier vs Developer Tier

| Feature | Free Tier | Developer Tier |
|---|---|---|
| **Cost** | Free (no payment method) | $5 free credit + pay-as-you-go |
| **Payment method** | Not required | Required (credit card via Stripe) |
| **Models** | All Production & Preview | All Production & Preview |
| **RPM (70B models)** | 40 | 240 (6x higher) |
| **RPM (8B models)** | 40 | 1,440 (36x higher) |
| **RPM (DeepSeek-R1)** | 20 | 60 (3x higher) |
| **RPD** | 40 per model | 6,000–288,000 per model |
| **TPD (tokens/day)** | 200,000 per model | No daily token cap |
| **Support** | Community | Community (email + forums) |

The Free Tier is extremely restrictive at just 40 requests per day. The Developer Tier is where real building happens.

---

## Available Models

### Production Models

| Model | Context | Input Price | Output Price |
|---|---|---|---|
| **Meta-Llama-3.1-8B-Instruct** | 16K | $0.01/M tokens | $0.02/M tokens |
| **Meta-Llama-3.3-70B-Instruct** | 128K | $0.06/M tokens | $0.12/M tokens |
| **DeepSeek-R1-Distill-Llama-70B** | 128K | $0.07/M tokens | $0.14/M tokens |
| **DeepSeek-V3.1** | 128K | $0.015/M tokens | $0.075/M tokens |
| **DeepSeek-V3-0324** | 128K | $0.30/M tokens | $0.45/M tokens |
| **DeepSeek-R1-0528** | 128K | $0.50/M tokens | $0.70/M tokens |

### Preview Models (evaluation only, may be removed without notice)

| Model | Context | Input Price | Output Price |
|---|---|---|---|
| **Llama-4-Maverick-17B-128E-Instruct** | 128K | $0.063/M tokens | $0.18/M tokens |
| **gpt-oss-120b** | 128K | $0.022/M tokens | $0.059/M tokens |
| **Qwen3-32B** | 8K | $0.04/M tokens | $0.08/M tokens |
| **Qwen3-235B** | — | $0.04/M tokens | $0.08/M tokens |
| **Whisper-Large-v3** (audio) | — | $10/hour | — |
| **E5-Mistral-7B-Instruct** (embeddings) | 4K | $0.013/M tokens | Free |
| **Llama-3.3-Swallow-70B-Instruct-v0.4** | 16K | — | — |

SambaNova is a **Llama 4 launch partner** — Maverick is available in the playground and via API.

---

## Rate Limits (Developer Tier)

### Production Models

| Model | RPM | RPD |
|---|---|---|
| Meta-Llama-3.1-8B-Instruct | 1,440 | 288,000 |
| Meta-Llama-3.3-70B-Instruct | 240 | 48,000 |
| DeepSeek-R1-Distill-Llama-70B | 240 | 48,000 |
| DeepSeek-V3-0324 | 60 | 12,000 |
| DeepSeek-V3.1 | 60 | 12,000 |
| DeepSeek-R1 | 60 | 12,000 |

### Preview Models

| Model | RPM | RPD |
|---|---|---|
| Whisper-Large-v3 | 450 | 90,000 |
| Llama-4-Maverick-17B-128E | 60 | 12,000 |
| gpt-oss-120b | 60 | 12,000 |
| E5-Mistral-7B-Instruct | 60 | 12,000 |
| Llama-3.3-Swallow-70B | 60 | 12,000 |
| Qwen3-32B | 30 | 6,000 |

---

## How Far $5 Goes

With $5 in credits, here is roughly what you can expect:

| Model | Approx. Tokens (input) |
|---|---|
| Llama 3.1 8B | ~500M tokens ($0.01/M) |
| Llama 3.3 70B | ~83M tokens ($0.06/M) |
| DeepSeek-V3.1 | ~333M tokens ($0.015/M) |
| DeepSeek-R1-Distill-Llama-70B | ~71M tokens ($0.07/M) |
| DeepSeek-R1-0528 | ~10M tokens ($0.50/M) |

The cheapest models give you hundreds of millions of tokens — enough for serious prototyping.

---

## Credit Expiration

There is some inconsistency between official sources:
- The **blog post** (July 2025 launch) states credits expire in **3 months**
- The **plans page** currently states credits expire in **30 days**

The policy may have been updated since launch. Check the [plans page](https://cloud.sambanova.ai/plans) for the current expiration window when you sign up. Either way, set a reminder — unused credits disappear.

---

## What Happens After Credits Run Out

1. **If you have a payment method on file:** You switch to **pay-as-you-go** billing. Usage is tracked per model and billed monthly via Stripe
2. **If you have no payment method:** You fall back to the **Free Tier** with severely limited rate limits (40 RPD, 200K TPD per model)
3. **No automatic charges** will surprise you if you never added a payment method

---

## API Integration

SambaNova uses an **OpenAI-compatible API**, so switching from OpenAI or other providers is trivial:

```python
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_SAMBANOVA_API_KEY",
    base_url="https://api.sambanova.ai/v1"
)

response = client.chat.completions.create(
    model="Meta-Llama-3.3-70B-Instruct",
    messages=[{"role": "user", "content": "Hello!"}]
)
```

Works with the OpenAI Python SDK, JavaScript SDK, LiteLLM, and any OpenAI-compatible client.

---

## Additional Tips

- **Speed advantage** — SambaNova runs on custom RDU chips, not GPUs. Expect inference speeds of **100+ tokens/second** on large models and **600+ tokens/second** on small models like Llama 8B. This is the main reason to choose SambaNova over competitors
- **AWS Marketplace** — you can also subscribe via [AWS Marketplace](https://aws.amazon.com/marketplace/pp/prodview-et56bohw7qcx2) and pay through your AWS bill instead of Stripe
- **Community credits** — join the [SambaNova Developer Community](https://community.sambanova.ai/) and participate in the leaderboard. Top 10 contributors each month earn **$10 in additional credits**
- **No free tier anymore** — SambaNova has confirmed they have no plans to maintain a permanent free tier. The Free Tier (no payment method) still works but with very low limits (40 RPD). The Developer Tier with $5 credit is the new default
- **Preview models** can be removed at any time without notice — do not build production workflows on them
- **Enterprise Tier** — if you need custom rate limits, dedicated nodes, or bring-your-own-checkpoints (BYOC), contact SambaNova sales

---

**Sources:**
- [SambaNova Cloud Plans](https://cloud.sambanova.ai/plans)
- [SambaNova Developer Tier Blog Post](https://sambanova.ai/blog/sambanova-cloud-developer-tier-is-live)
- [SambaNova Rate Limits Documentation](https://docs.sambanova.ai/docs/en/models/rate-limits)
- [SambaNova Supported Models](https://docs.sambanova.ai/docs/en/models/sambacloud-models)
- [SambaNova Cloud Pricing](https://cloud.sambanova.ai/plans/pricing)
- [SambaNova Quickstart Guide](https://docs.sambanova.ai/docs/en/get-started/quickstart)
- [SambaNova on AWS Marketplace](https://aws.amazon.com/marketplace/pp/prodview-et56bohw7qcx2)
- [SambaNova Community Discussion on Free Tier](https://community.sambanova.ai/t/is-free-tier-going-away/847)

## Submission 68: Amazon SageMaker Studio Lab - Free ML Environment

- Category: Free GPU & Compute
- Source Link: https://studiolab.sagemaker.aws/
- Amount: 4 hours GPU/day + 8 hours CPU/day; 15GB persistent storage
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Amazon SageMaker Studio Lab is a **completely free**, browser-based ML development environment powered by JupyterLab 4. It provides access to an **NVIDIA T4 GPU** (4 hours/day) and a **T3.xlarge CPU** (8 hours/day), **15 GB of persistent storage**, and pre-installed ML frameworks -- all without needing an AWS account or credit card. It is one of the best zero-cost options for learning, prototyping, and experimenting with machine learning.

---

## Registration (Step-by-Step)

1. Go to [studiolab.sagemaker.aws](https://studiolab.sagemaker.aws/) and click **"Request free account"**
2. Fill in the request form: **email address**, first/last name, country, organization name, and occupation
3. Click **"Submit request"** -- you will receive an email to verify your email address; click the verification link
4. **Wait for approval** -- AWS reviews requests within **5 business days**. In practice, most approvals come within a few hours to a few days
5. Once approved, you receive an email with a **registration link** -- you have **7 days to claim your account** before the link expires
6. Click the link and create your account: choose a **username** and **password** (this is separate from any AWS account)
7. Verify your email one more time via the confirmation email
8. **First runtime launch -- phone verification (one-time):** Enter a mobile phone number, receive a **6-digit SMS code**, and verify. This is required only once
9. Choose **CPU** or **GPU** runtime and click **"Start runtime"** -- your JupyterLab environment loads in the browser

**Important:**
- No AWS account or credit card is required at any step
- One account per person/email
- If your approval link expires after 7 days, you must submit a new request
- Phone verification uses the AWS SMS channel, which supports **240+ countries** but has known delivery issues in some regions (China, Colombia, UAE, Jordan have been reported). VoIP numbers typically do not work
- Referral codes (used for workshops and hackathons) bypass the approval wait and grant instant access

---

## Compute Resources & Limits

| Resource | Instance Type | Session Limit | Daily Limit |
|---|---|---|---|
| **GPU** | G4dn.xlarge (NVIDIA T4, 16 GB VRAM) | 4 hours/session | 4 hours per 24-hour period |
| **CPU** | T3.xlarge (4 vCPUs, 16 GB RAM) | 4 hours/session | 8 hours per 24-hour period |

**Key details:**
- Only **one runtime session** can be active at a time (you cannot run CPU and GPU simultaneously)
- When your session time runs out, **all running computations stop** -- but your files and installed packages are saved to persistent storage
- Compute availability is **not guaranteed** and is subject to demand. During peak times, you may not be able to start a GPU session immediately
- Time limit increases are **not supported**
- You can switch between CPU and GPU runtimes between sessions

**Note:** The GPU provides an NVIDIA T4 with 16 GB of GPU memory, which is sufficient for training small-to-medium deep learning models, fine-tuning pre-trained models, and running inference on models like Stable Diffusion or smaller LLMs.

---

## Storage

- **15 GB of persistent storage** per user
- Files, notebooks, conda environments, and installed packages **persist across sessions and reboots**
- Environment state is automatically saved when you update packages or create new files
- **Manual save recommended:** File edits are periodically auto-saved during a session, but are **not saved when the runtime ends**. Always save your work manually (Ctrl+S) before your session expires
- No option to expand storage beyond 15 GB

**Storage tips:**
- Avoid storing large datasets directly -- link to external sources (S3, Hugging Face Hub, GitHub) and load data in batches
- Large model checkpoints and pre-trained weights can fill 15 GB quickly. Use model streaming or download only what you need per session
- Use `.gitignore` to keep your Git repos clean and avoid syncing large binary files

---

## Pre-installed Frameworks & Environment

Studio Lab comes with a **default conda environment** and also supports the **SageMaker Distribution**, a curated environment designed to match full SageMaker Studio:

| Framework / Library | Included |
|---|---|
| **PyTorch** | Yes |
| **TensorFlow** | Yes |
| **Keras** | Yes |
| **NumPy** | Yes |
| **scikit-learn** | Yes |
| **Pandas** | Yes |
| **Hugging Face Transformers** | Installable via pip/conda |
| **OpenCV** | Installable via pip/conda |
| **XGBoost** | Installable via pip/conda |

- **JupyterLab 4** with full extension support
- **Python 3.9+** (default)
- Package managers: **conda**, **pip**, and **micromamba** are all supported
- Custom conda environments persist across sessions
- Custom JupyterLab extensions persist across sessions
- **Git integration** built-in -- clone repos, push/pull, manage branches directly from the UI

---

## Comparison with Alternatives

| Feature | SageMaker Studio Lab | Google Colab (Free) | Kaggle Notebooks |
|---|---|---|---|
| **Free GPU** | NVIDIA T4 (guaranteed type) | T4 (not guaranteed) | T4 or P100 |
| **GPU time/day** | 4 hours | ~4-12 hours (variable) | 30 hours/week |
| **Persistent storage** | 15 GB (persistent) | None (uses Google Drive) | None (session-only) |
| **Sign-up** | Approval required (1-5 days) | Google account (instant) | Kaggle account (instant) |
| **Environment** | JupyterLab 4 (fully customizable) | Colab notebook (limited) | Kaggle notebook |
| **Package persistence** | Yes | No | No |
| **Credit card needed** | No | No | No |

**Key advantages of Studio Lab:**
- **Persistent storage and environments** -- your installed packages, datasets, and configurations survive between sessions, unlike Colab where you reinstall everything each time
- **Guaranteed GPU type** -- you always get a T4, whereas Colab may downgrade you to a weaker GPU or deny access entirely
- **Full JupyterLab** -- not a simplified notebook interface. You get a file browser, terminal, multiple tabs, and extension support

**Key disadvantages:**
- **Approval wait** -- unlike Colab or Kaggle, you cannot start immediately
- **Shorter GPU sessions** -- 4 hours/day vs. Colab's variable (but often longer) sessions
- **No real-time collaboration** -- unlike Colab, you cannot share notebooks for live co-editing

---

## Migration to Full SageMaker

Studio Lab is designed as an on-ramp to the full **Amazon SageMaker** platform:

1. Export your Studio Lab notebooks and environment files
2. Create an AWS account (requires credit card)
3. Open **SageMaker Studio** in the AWS Console
4. Import your notebooks and recreate your conda environment using the SageMaker Distribution
5. Access larger instances, SageMaker Pipelines (CI/CD), model deployment, and other production features

**What Studio Lab does NOT include** (available in full SageMaker):
- SageMaker Pipelines for ML CI/CD
- Real-time model endpoints / inference
- SageMaker GroundTruth (data labeling)
- Built-in SageMaker algorithms and estimators
- Fine-grained IAM access control
- Configurable instance types and storage

---

## Tips for Getting the Most Out of Studio Lab

- **Use checkpoints for GPU training** -- save model state periodically so you can resume in the next 4-hour session rather than starting over
- **Prefer the CPU runtime for data preprocessing** -- save your GPU hours for actual training. CPU sessions give you 8 hours/day
- **Install packages once, use them forever** -- packages installed via conda/pip persist across sessions. No need to reinstall like in Colab
- **Clone GitHub repos for reproducibility** -- the built-in Git integration makes it easy to version your work
- **Use the SageMaker Distribution environment** if you plan to eventually migrate to full SageMaker -- it ensures compatibility
- **Keep an eye on storage** -- 15 GB fills up fast with model weights. Clean up old checkpoints and datasets regularly
- **If GPU is unavailable**, try again during off-peak hours (late night / early morning US time). Availability depends on demand
- **For workshops or classrooms**, request referral codes from AWS to give participants instant access without the approval wait

---

## Related Free Programs

- **AWS Free Tier** -- new AWS accounts get up to $200 in credits (post-July 2025), usable for SageMaker and Bedrock. Requires credit card
- **AWS Educate** -- additional $200 in credits for students. Requires .edu email
- **AWS Activate** -- up to $100K in credits for startups
- **Google Colab** -- alternative free GPU environment with instant access but no persistent storage
- **Kaggle Notebooks** -- 30 GPU hours/week, instant access, great dataset ecosystem
- **Lightning AI Studios** -- free tier with 22 GPU hours/month on T4

---

**Sources:**
- [SageMaker Studio Lab Homepage](https://studiolab.sagemaker.aws/)
- [SageMaker Studio Lab FAQ](https://studiolab.sagemaker.aws/faq)
- [AWS Docs: Studio Lab Overview](https://docs.aws.amazon.com/sagemaker/latest/dg/studio-lab-overview.html)
- [AWS Docs: Onboard to Studio Lab](https://docs.aws.amazon.com/sagemaker/latest/dg/studio-lab-onboard.html)
- [AWS Blog: SageMaker Studio Lab Preview Announcement](https://aws.amazon.com/blogs/aws/now-in-preview-amazon-sagemaker-studio-lab-a-free-service-to-learn-and-experiment-with-ml/)
- [DataCamp: SageMaker Studio Lab Guide](https://www.datacamp.com/tutorial/sagemaker-studio-lab)
- [Roboflow: SageMaker Studio Lab vs Google Colab](https://blog.roboflow.com/sagemaker-studio-lab-vs-google-colab/)
- [GitHub: SageMaker Distribution](https://github.com/aws/sagemaker-distribution)
- [GitHub: Studio Lab Examples](https://github.com/aws/studio-lab-examples)

## Submission 67: RunPod New User Bonus Credits - $5-$500

- Category: Free GPU & Compute
- Source Link: https://www.runpod.io/
- Amount: $5-$500 random bonus (after first $10 spend)
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

New RunPod users receive a **random credit bonus between $5 and $500** after their first $10 spend on the platform. RunPod offers per-second billing for GPU compute across Community Cloud (peer-hosted, cheaper) and Secure Cloud (enterprise data centers). No long-term contracts required — pure pay-as-you-go. Credits cannot be withdrawn as cash.

---

## Registration (Step-by-Step)

1. Go to [runpod.io](https://www.runpod.io) and click **Sign Up**
2. Create an account with email or sign in with Google/GitHub
3. Add a **payment method** (credit card or crypto deposit) — required before deploying any GPU
4. Add at least **$10** to your RunPod balance (prepaid credits model)
5. After your first $10 spend, you automatically receive a **random bonus between $5 and $500**
6. Bonus credits appear in your account and can be used for any RunPod service (Pods, Serverless, storage)

**Important:**
- The bonus is awarded after you **spend** $10, not just deposit it
- One bonus per new user account
- Bonus credits cannot be withdrawn — they can only be used on RunPod services
- RunPod uses a prepaid credits model — you load funds first, then deploy

---

## GPU Pod Pricing (On-Demand, Per Hour)

RunPod bills by the **millisecond** but displays rates per hour. Two cloud tiers are available:

- **Community Cloud** — GPUs hosted by vetted third-party providers. 20-30% cheaper than Secure Cloud. Good for experimentation, training, and non-critical workloads
- **Secure Cloud** — enterprise-grade T3/T4 data centers with SOC2 compliance. Higher reliability and persistent NVMe-backed network volumes. Better for production

### Representative GPU Pricing

| GPU | VRAM | Community Cloud | Secure Cloud |
|---|---|---|---|
| **RTX 4090** | 24 GB | ~$0.34-$0.39/hr | ~$0.44-$0.69/hr |
| **L4** | 24 GB | ~$0.24/hr | ~$0.34/hr |
| **RTX 3090** | 24 GB | ~$0.22/hr | — |
| **A40** | 48 GB | ~$0.59/hr | ~$0.69/hr |
| **L40S** | 48 GB | ~$0.74/hr | ~$0.84/hr |
| **RTX 6000 Ada** | 48 GB | ~$0.71/hr | ~$0.81/hr |
| **A100 SXM** | 80 GB | ~$1.79-$1.89/hr | ~$1.99/hr |
| **H100 SXM** | 80 GB | ~$1.99-$2.99/hr | ~$2.69/hr |
| **H200 SXM** | 141 GB | Contact sales | ~$4.31/hr |
| **B200** | 180 GB | Contact sales | Contact sales |

*Prices fluctuate based on supply and demand. Check [runpod.io/pricing](https://www.runpod.io/pricing) for current rates.*

### Spot Instances

Spot (interruptible) instances are available at **50-70% lower cost** than on-demand. Suitable for fault-tolerant training jobs and batch inference. Pods may be interrupted with short notice when demand rises.

### Savings Plans

- **1-month commitment** — 15% discount on Secure Cloud
- **3-month commitment** — 20% discount on Secure Cloud
- Requires upfront payment for the term

---

## Serverless Endpoints

RunPod Serverless lets you deploy models behind auto-scaling API endpoints without managing infrastructure. Two worker types:

- **Flex Workers** — scale to zero when idle, pay only during processing. Best for variable/bursty workloads
- **Active Workers** — always-on with 20-30% discount vs. flex. Best for latency-sensitive production workloads

### Serverless GPU Pricing

| GPU | VRAM | Flex (per hr) | Active (per hr) |
|---|---|---|---|
| **A4000 / RTX 4000** | 16 GB | $0.576 | $0.396 |
| **L4 / A5000 / RTX 3090** | 24 GB | $0.684 | $0.468 |
| **RTX 4090 PRO** | 24 GB | $1.116 | $0.756 |
| **A6000 / A40** | 48 GB | $1.224 | $0.864 |
| **L40 / L40S / 6000 Ada PRO** | 48 GB | $1.908 | $1.332 |
| **A100** | 80 GB | $2.736 | $2.160 |
| **H100 PRO** | 80 GB | $4.176 | $3.348 |
| **H200 PRO** | 141 GB | $5.580 | $4.464 |
| **B200** | 180 GB | $8.640 | $6.840 |

Serverless billing includes **start time** (model loading), **execution time** (processing), and configurable **idle time** (worker stays warm). Billed per second.

---

## Storage Pricing

| Storage Type | Running Pods | Stopped Pods |
|---|---|---|
| **Container Disk** | $0.10/GB/month | N/A (destroyed on stop) |
| **Volume (Pod storage)** | $0.10/GB/month | $0.20/GB/month |
| **Network Volume (<1 TB)** | $0.07/GB/month | $0.07/GB/month |
| **Network Volume (>1 TB)** | $0.05/GB/month | $0.05/GB/month |

**No ingress or egress fees** — data transfer in and out is free.

---

## Pre-Built Templates

RunPod provides one-click templates for popular AI workloads:

- **Stable Diffusion** (Automatic1111, ComfyUI)
- **LLM inference** (vLLM, Text Generation Inference)
- **Jupyter Notebook** with PyTorch/CUDA pre-installed
- **Kohya SS** for LoRA training
- Custom Docker containers supported

---

## Startup Program

RunPod offers a separate **Startup Program** with significantly larger credit grants:

- Up to **1,000 hours of free H100 GPU time**
- Up to **1,000,000 free serverless GPU requests**
- Credit amount based on your projected GPU usage and planned spend
- Apply at [runpod.io/startup-program](https://www.runpod.io/startup-program)
- For research labs and academic teams, contact **startups@runpod.io** for case-by-case consideration

This is separate from the new user bonus and requires an application.

---

## Referral Program

After signing up, you can earn additional credits through RunPod's referral program:

- Both referrer and referred user get a **$5-$500 random bonus** (after the referred user spends $10)
- Referrer also earns **3% commission on Pod spend** and **5% on Serverless spend** for the referred user's first 6 months

---

## Additional Tips

- **Your $10 goes further than you think** — an RTX 4090 at $0.34/hr means $10 buys roughly 29 hours of GPU time. With the bonus credits on top, you could get 40+ hours for your first $10
- **Use Spot instances for training** — save 50-70% on non-critical workloads. Set up checkpointing so interrupted jobs can resume
- **Community Cloud is fine for experimentation** — save 20-30% vs. Secure Cloud when uptime guarantees don't matter
- **Serverless for inference APIs** — if you're deploying a model for users, serverless flex workers scale to zero and you pay nothing when idle
- **No egress fees** — unlike AWS/GCP/Azure, RunPod does not charge for data transfer, which can save significantly on large model downloads
- **Pre-load models on Network Volumes** — persistent storage lets you avoid re-downloading models each time you start a Pod
- **RunPod vs. major clouds** — RunPod's H100 at ~$2.79/hr compares favorably to AWS at ~$12.29/hr (77% cheaper) for equivalent hardware

---

**Sources:**
- [RunPod Pricing](https://www.runpod.io/pricing)
- [RunPod Serverless Pricing Documentation](https://docs.runpod.io/serverless/pricing)
- [RunPod Pod Pricing Documentation](https://docs.runpod.io/pods/pricing)
- [RunPod Startup Program](https://www.runpod.io/startup-program)
- [RunPod Referral Program](https://www.runpod.io/referral-and-affiliate-program)
- [RunPod GPU Cloud Product Page](https://www.runpod.io/product/cloud-gpus)
- [RunPod Serverless Product Page](https://www.runpod.io/product/serverless)

## Submission 66: Replicate - Free Generations for New Users

- Category: AI API Free Tiers
- Source Link: https://replicate.com/
- Amount: Limited free runs on select models (exact number not publicly disclosed), then pay-as-you-go
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Replicate gives new users a **limited number of free runs** on select AI models with **no credit card required** — just sign up with GitHub, Google, or email and start generating. The free tier covers popular image, video, and utility models from their "Try for Free" collection. After your free runs are exhausted, you switch to pay-as-you-go billing charged per second of compute time. Replicate hosts **thousands of open-source and proprietary models** for text, image, audio, and video generation.

**Note:** Replicate does **not** publicly disclose the exact number of free runs (the "50 free generations/month" figure sometimes cited is not confirmed by Replicate's official documentation). The free tier is a one-time allowance for experimentation, not a monthly reset.

---

## Registration (Step-by-Step)

1. Go to [replicate.com](https://replicate.com)
2. Click **"Sign in"** in the top-right corner
3. Choose your sign-in method: **GitHub**, **Google**, or **email** — no credit card required
4. If using GitHub or Google, authorize the OAuth connection; if using email, confirm via the verification email
5. Once logged in, you land on your dashboard with access to the model explorer and your API token
6. To get your **API token**: click your avatar in the top-right, go to **API Tokens**, and copy or create a new token (tokens start with `r8_`)
7. Start running models from the web playground or via the API — free runs are available immediately on select models

**Important:**
- No credit card or payment method is needed to sign up and use the free tier
- A payment method (credit card, debit card, or bank transfer) is only required when you exceed the free allowance
- One account per person — standard GitHub/Google/email identity

---

## Free "Try for Free" Models

Replicate curates a [Try for Free collection](https://replicate.com/collections/try-for-free) of models you can run without setting up billing. Notable models include:

### Image Generation

| Model | Description |
|---|---|
| **Google Imagen 4** | Google's latest text-to-image model, high quality and easy to prompt |
| **FLUX Kontext Pro** | Strong prompt adherence with style control for photoreal and illustrated outputs |
| **FLUX 1.1 Pro** | Improved FLUX model with consistent image quality and diversity |
| **FLUX Dev** | Experimental 12B parameter model, good for testing edge cases |
| **Ideogram V3 Turbo** | Fast creative generation, good for posters, products, and text in images |

### Video Generation

| Model | Description |
|---|---|
| **Minimax Video-01** | Generate videos from text or images, with consistent characters |
| **Luma Reframe Video** | Reframe and resize videos up to 30 seconds for social posts |
| **Topaz Video Upscale** | Enhance and upscale low-resolution videos |

### Image Restoration & Upscaling

| Model | Description |
|---|---|
| **Topaz Image Upscale** | Professional-grade image upscaling with minimal artifacts |
| **CodeFormer** | Restore low-quality or AI-generated faces |
| **GFPGAN** | Fast face restoration for old or damaged photos |

The free collection changes over time — new models are added and older ones may rotate out.

---

## Full Model Catalog (Paid)

Beyond the free collection, Replicate hosts thousands of models across multiple categories:

- **Language Models (LLMs):** GPT-5, Claude 4.5 Sonnet, Gemini 2.5 Flash, Llama 3.1 405B, DeepSeek R1, and many more
- **Image Generation:** FLUX Pro, Stable Diffusion, Ideogram, and hundreds of community fine-tunes
- **Video Generation:** Minimax, Luma, and emerging video models
- **Audio & Speech:** Text-to-speech, speech-to-text, and music generation models
- **Utility Models:** Background removal, upscaling, face restoration, object detection

All models are accessible via a unified REST API and official client libraries (Python, Node.js, Swift, etc.).

---

## Pricing After Free Tier

Replicate uses **pay-as-you-go** billing. Some models charge per second of compute time, others per token or per output.

### Hardware Pricing (per-second billing)

| Hardware | Cost/sec | Cost/hr |
|---|---|---|
| CPU (Small) | $0.000025 | $0.09 |
| CPU | $0.0001 | $0.36 |
| Nvidia T4 GPU | $0.000225 | $0.81 |
| Nvidia L40S GPU | $0.000975 | $3.51 |
| Nvidia A100 (80GB) | $0.0014 | $5.04 |
| Nvidia H100 | $0.001525 | $5.49 |

### Example Model Pricing

| Model | Price |
|---|---|
| **FLUX 1.1 Pro** | ~$0.04 per image |
| **Ideogram V3** | ~$0.09 per image |
| **Claude 3.7 Sonnet** | $3.00/M input tokens, $15.00/M output tokens |

### Billing Options

- **Prepaid credit:** Purchase credit upfront, usage deducts from balance
- **Arrears billing:** Charged at the start of each month for the previous month's usage
- Failed runs are **never charged**

---

## Fine-Tuning & Custom Models

Replicate supports fine-tuning image models (especially FLUX) directly from the web or via API:

- **No ML expertise required** — upload a handful of images and set a trigger word
- **Cost:** ~$1.85 for a typical 20-minute training on H100 hardware (20 images, 1000 steps). The fast FLUX trainer does it in under 2 minutes for under $2
- **Custom models via Cog:** Deploy your own ML models using Replicate's open-source [Cog](https://github.com/replicate/cog) packaging tool
- Fine-tuning requires billing to be set up (not available on the free tier)

---

## Rate Limits

| Account Type | Limit |
|---|---|
| **Without billing** (free tier) | 1 request/sec, max 6 requests/min |
| **With billing** | 600 predictions/min, 3,000 other API calls/min |

As your credit balance approaches zero, rate limits tighten to prevent overspending. Contact Replicate for higher limits.

---

## Data Retention & Privacy

- All input parameters, output values, and logs are **automatically deleted after 1 hour** by default for API predictions
- You must save outputs yourself if you want to keep them
- Webhook delivery is available for async prediction results

---

## Additional Tips

- **Commercial use** depends on each model's license — always check the license on the model card before using outputs commercially. Even free-to-run models may have restrictions
- **No monthly free tier reset** — the free allowance is for initial experimentation, not a recurring monthly benefit
- **API tokens** are 40-character strings starting with `r8_` — treat them like passwords. Use environment variables, never hardcode them
- **Web playground** lets you test any model without writing code — great for evaluating before committing to API integration
- **Replicate vs. Hugging Face** — Replicate focuses on hosted inference with zero setup, while Hugging Face offers more model discovery and self-hosting options. Replicate is better for quick API integration; Hugging Face for research and local deployment
- **No enterprise plan required** — everything is pay-as-you-go. Enterprise users can negotiate committed spend for multi-GPU access and volume discounts

---

**Sources:**
- [Replicate Pricing](https://replicate.com/pricing)
- [Replicate Billing Documentation](https://replicate.com/docs/topics/billing)
- [Try for Free Collection](https://replicate.com/collections/try-for-free)
- [Replicate Rate Limits](https://replicate.com/docs/topics/predictions/rate-limits)
- [Replicate Language Models Collection](https://replicate.com/collections/language-models)
- [Replicate Explore Page](https://replicate.com/explore)
- [Fine-Tune FLUX on Replicate](https://replicate.com/blog/fine-tune-flux)
- [Getting Started with Python](https://replicate.com/docs/get-started/python)
- [API Tokens Documentation](https://replicate.com/docs/topics/security/api-tokens)

## Submission 65: Perplexity Pro API Credit - $5/Month for Subscribers

- Category: AI API Free Tiers
- Source Link: https://www.perplexity.ai/settings/api
- Amount: $5/month in API credits (with Pro subscription); $50/month with Max subscription
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Perplexity Pro subscribers ($20/month) receive **$5 in monthly API credits** on the first day of each month. The API provides access to the **Sonar model family** — search-augmented LLMs that combine web retrieval with language generation. Credits do not roll over; unused balance resets monthly. Students and educators can get free Pro access (including the $5/month API credit) for up to **12 months** through the Education Plan, verified via SheerID.

---

## Registration (Step-by-Step)

### For Pro Subscribers

1. Go to [perplexity.ai](https://www.perplexity.ai) and create an account (email, Google, or Apple sign-in)
2. Subscribe to **Perplexity Pro** ($20/month or $200/year) — credit card required
3. Navigate to [perplexity.ai/settings/api](https://www.perplexity.ai/settings/api)
4. Your **$5 monthly credit** should appear automatically within 10-20 minutes of subscribing
5. Click **"Generate"** to create your API key (starts with `pplx-`)
6. Copy and save the key immediately — it is shown only once

**Important:**
- Do **not** manually add $5 credits yourself — this will result in a $5 charge on top of the complimentary credit
- The API is billed separately from the Pro chat subscription. "Unlimited" queries in the Pro plan apply only to the web/app interface, not the API
- Pro subscribers get $5/month; Max subscribers ($200/month) get $50/month in API credits

### For Students / Educators (Free Pro)

1. Go to [perplexity.ai/students](https://www.perplexity.ai/students) or click **Upgrade** in the app and toggle to the **Education** plan
2. Click **"Verify as student"** — this uses [SheerID](https://www.sheerid.com) for verification
3. Enter your **.edu** or **.ac** email address (personal Gmail/Yahoo will not work)
4. Most academic emails are verified **instantly**. If not, upload proof (student ID, enrollment letter) — manual review takes 24-48 hours
5. Once verified, you get **Pro access free for 1 month** with the option to extend via the referral program
6. Navigate to API settings and generate your key as described above

**Student referral program:** Refer classmates to earn additional free months of Pro, stackable up to **24 months total**. The referral program ends May 31, 2026.

---

## Available API Models (Sonar Family)

Perplexity does not expose third-party models (GPT, Claude, etc.) through its API. Instead, it provides its own **Sonar** model family — search-augmented LLMs with built-in web retrieval.

| Model | Model ID | Input (per 1M tokens) | Output (per 1M tokens) | Context Window | Best For |
|---|---|---|---|---|---|
| **Sonar** | `sonar` | $1 | $1 | 127K | Quick searches, fact-checking, simple Q&A |
| **Sonar Pro** | `sonar-pro` | $3 | $15 | 200K | Multi-step research, complex queries, 2x more citations |
| **Sonar Reasoning Pro** | `sonar-reasoning-pro` | $2 | $8 | — | Chain-of-thought reasoning with web search |
| **Sonar Deep Research** | `sonar-deep-research` | $2 | $8 | 128K | Comprehensive reports, multi-source synthesis |

**Additional costs:**
- **Search API** (raw web results, no synthesis): $5 per 1,000 requests
- **Search context pricing** varies by depth: Low ($5-6/K requests), Medium ($8-10/K), High ($12-14/K)
- **Citation tokens** are no longer billed for Sonar and Sonar Pro as of 2026
- **Reasoning tokens** (Deep Research): $3 per 1M tokens

---

## How Far Does $5 Go?

With the base **Sonar** model at $1/M input + $1/M output tokens:

| Scenario | Approx. Queries with $5 |
|---|---|
| Short Q&A (~500 input + 500 output tokens) | ~5,000 queries |
| Medium search (~1K input + 2K output tokens) | ~1,600 queries |
| Long research (~2K input + 4K output tokens) | ~830 queries |

With **Sonar Pro** at $3/M input + $15/M output tokens, $5 covers roughly **250-300 medium queries** — significantly fewer. Choose your model based on complexity needs.

---

## Credit & Billing Details

- **$5 credit resets** on the 1st of each month — unused credits do **not** roll over
- **Top-up credits** (purchased separately) do persist in your balance until used
- Beyond the $5 credit, usage is pay-as-you-go charged to your payment method
- **Refunds:** Available within 14 days of purchase if credits are unused. Used credits are non-refundable
- You can set up **auto top-off** to automatically add credits when your balance drops below a threshold
- API billing is entirely separate from the Pro subscription fee

---

## Subscription Plans Comparison

| Plan | Price | Monthly API Credit | Chat Queries |
|---|---|---|---|
| **Free** | $0 | None (pay-as-you-go only) | Limited |
| **Pro** | $20/mo ($200/yr) | **$5/month** | Unlimited concise + Pro |
| **Max** | $200/mo ($2,000/yr) | **$50/month** | Unlimited + advanced models |
| **Education Pro** | Free for 1-24 months (then $4.99/mo) | **$5/month** (while on Pro) | Same as Pro |
| **Enterprise Pro** | $40/seat/mo | $5/seat/month | 500 research/day |

---

## Education Plan Details

- **Eligibility:** Current students or faculty at university-level institutions or higher
- **Verification:** Via [SheerID](https://www.sheerid.com) using a .edu or .ac email domain
- **Duration:** 1 month free initially, extendable to 24 months via referrals
- **After free period:** Downgrades to free tier automatically (no charges). Students can then subscribe at the discounted rate of **$4.99/month**
- **Features included:** Full Pro access — Study Mode, Perplexity Labs, unlimited uploads, access to advanced models in the chat interface
- **No credit card required** for the free Education Pro signup

---

## Additional Tips

- **API vs. Chat are separate products** — the Pro subscription gives you unlimited chat queries on the website/app, but the API has its own token-based billing. The $5 credit is the only bridge between them
- **OpenAI-compatible endpoint** — the Perplexity API uses OpenAI-compatible chat completions format (`/chat/completions`), making it easy to integrate with existing tooling
- **Data privacy** — API (Sonar) queries are never logged or used for model training, unlike the free chat interface
- **No free API tier exists** — the free Perplexity plan does not include any API credits. You must either subscribe to Pro or purchase credits directly
- **Key rotation** — after creating your first key via the web UI, you can manage additional keys programmatically via the API
- **Rate limits** — API rate limits depend on your credit balance and plan tier. Higher balances unlock higher concurrency
- **Alternative access** — Perplexity Sonar models are also available through [OpenRouter](https://openrouter.ai/perplexity), which may offer different pricing or bundling options

---

**Sources:**
- [Perplexity API Settings](https://www.perplexity.ai/settings/api)
- [Perplexity API Documentation](https://docs.perplexity.ai/)
- [Perplexity API Pricing](https://docs.perplexity.ai/getting-started/pricing)
- [Perplexity API Payment & Billing Help](https://www.perplexity.ai/help-center/en/articles/10354847-api-payment-and-billing)
- [Perplexity API Key Management](https://docs.perplexity.ai/guides/api-key-management)
- [Perplexity Subscription Plans](https://www.perplexity.ai/help-center/en/articles/11187416-which-perplexity-subscription-plan-is-right-for-you)
- [Perplexity Education Pro](https://www.perplexity.ai/help-center/en/articles/12590157-what-is-education-pro)
- [Perplexity Students Page](https://www.perplexity.ai/students)
- [SheerID Student Verification](https://shop.sheerid.com/offers/free-perplexity-pro-for-students/)
- [Perplexity Pricing Guide (Finout)](https://www.finout.io/blog/perplexity-pricing-in-2026)
- [Perplexity API Cost Guide (GlbGPT)](https://www.glbgpt.com/hub/perplexity-api-cost-2025/)

## Submission 64: Perplexity Education Plan - Free Pro for Students

- Category: Education & Student
- Source Link: https://www.perplexity.ai/education
- Amount: 12 months of Perplexity Pro ($240 value) including $5/month API credits. Extendable up to 18 months via referral program.
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Verified students and educators at accredited institutions get **12 months of Perplexity Pro completely free** — a $240 value. No credit card required. No auto-billing. After 12 months your account simply downgrades to the free tier. Includes unlimited Pro Search, advanced AI model access (GPT-5, Claude Sonnet 4.5, Gemini 3 Pro, Sonar), $5/month in API credits, file uploads, Deep Research, and education-exclusive features like Learn Mode (interactive flashcards and quizzes). Verification is handled through SheerID, which covers 264 million students across 190 countries.

---

## Registration (Step-by-Step)

1. Go to [perplexity.ai](https://www.perplexity.ai) and create an account (or sign in if you already have one)
2. Click the **"Upgrade"** button in the bottom-left corner of the app
3. Toggle to the **"Education"** plan tab
4. Click **"Verify as Student"** or **"Verify as Educator"**
5. You will be redirected to **SheerID verification** — enter your school name, full name, date of birth, and academic email address
6. Use your **.edu**, **.ac.uk**, or other recognized academic email domain — Gmail, Yahoo, and other personal email domains will not be accepted for verification
7. Most academic emails are **verified instantly** against SheerID's authoritative data sources. If instant verification fails, you will be prompted to upload a document (student ID, enrollment letter, etc.) — manual review typically completes within 3 minutes, but can take up to 24–48 hours in some cases
8. Once approved, your account is automatically upgraded to Education Pro — no payment information needed

**Important:**
- You do **not** need to use the same email for your Perplexity account and SheerID verification — they can be different
- Verification covers university-level institutions and higher (not high school)
- If your school is not recognized, contact Perplexity support for manual verification
- Desktop browser is recommended for the verification flow — the mobile experience can be inconsistent
- If you are still enrolled after 12 months, you can re-verify and get another 12 months free

---

## What's Included

### Pro Search & AI Models
- **Unlimited Pro Search** — multi-step reasoning, detailed research, complex queries
- **Multiple AI models** in one subscription: Perplexity Sonar, GPT-5.2, Claude Sonnet 4.5, Gemini 3 Pro
- **Reasoning models** for complex analytical tasks: Sonnet 4.5 Thinking, Grok 4
- **Best Mode** — auto-selects the optimal model for each query

### Education-Exclusive Features
- **Learn Mode** — generate interactive flashcards, summaries, and practice quizzes from your notes or study materials
- **Expanded Citations** — up to 10x more citations in answers compared to the free plan, ideal for academic research and writing papers

### Research & Productivity
- **Deep Research** — in-depth reports on complex topics with comprehensive source analysis
- **Unlimited file uploads** — PDFs, CSVs, audio, video, and images for contextual AI analysis
- **Perplexity Labs** — experimental AI tools, build dashboards, small apps, and structured reports
- **Image and video generation** with monthly quota

### API Access
- **$5/month in API credits** loaded on the first day of each month
- Access to **Sonar API models** (Sonar, Sonar Pro, Sonar Reasoning Pro) via pplx-api
- Generate API keys at Settings > API tab
- Credits do **not** roll over — unused balance resets monthly
- Once the $5 monthly credit is exhausted, additional API usage requires purchasing top-up credits (pay-as-you-go)
- API documentation at [docs.perplexity.ai](https://docs.perplexity.ai)

---

## API Models & Pricing

| Model | Best For | Notes |
|---|---|---|
| **Sonar** | Fast web search, quick retrieval | Lightweight, speed-optimized, most affordable |
| **Sonar Pro** | Deeper reasoning, larger context | Based on larger models, higher cost per token |
| **Sonar Reasoning Pro** | Complex analytical tasks | Premium tier, deepest analysis |

The API is pay-per-token after the $5 monthly credit. Citation tokens are no longer billed for standard Sonar and Sonar Pro models.

---

## Student Referral Program

Verified students can earn **additional free months** by referring classmates:
- Each friend who creates a new Perplexity account **and** verifies their student status via SheerID earns you **1 extra month** of Pro
- Maximum of **6 additional free months** through referrals (stacks on top of the initial 12 months)
- Total possible free access: **up to 18 months** (12 base + 6 referral)
- Find your unique referral link in the Perplexity app under your account settings

---

## What Happens After 12 Months

1. **No auto-billing** — your account simply downgrades to the free tier. No credit card was ever required, so there is nothing to charge
2. **Data and history preserved** — you keep your conversation history, just lose Pro features
3. **Re-verification available** — if you are still a student or educator, you can verify again for another 12 months
4. **Discounted continuation** — verified students and educators can subscribe to Education Pro at **$4.99/month** (75% off the standard $20/month Pro price) if they prefer not to re-verify or are no longer eligible

---

## Limitations & Gotchas

- **University-level only** — high school and K-12 students are not eligible
- **Personal email domains rejected** — Gmail, Yahoo, Outlook, etc. will not pass SheerID verification. You need an institutional email domain
- **API credits are minimal** — $5/month is enough for light experimentation, not production workloads
- **No Max plan features** — Education Pro matches the $20/month Pro plan, not the higher-tier Max plan. You do not get unlimited Labs access, Research mode without caps, or priority support
- **SheerID verification can be slow** — while most verifications are instant, edge cases (unrecognized institutions, manual document review) can take 1–2 days
- **Referral program may expire** — the student referral program has specific end dates (check current terms)

---

## Free Plan vs Education Pro Comparison

| Feature | Free Plan | Education Pro |
|---|---|---|
| Basic searches | Unlimited | Unlimited |
| Pro Search | Limited daily | Unlimited |
| AI models | Default only | GPT-5.2, Claude 4.5, Gemini 3 Pro, Sonar |
| File uploads | Limited | Unlimited |
| Learn Mode | No | Yes |
| Expanded citations | No | Up to 10x more |
| Deep Research | No | Yes |
| API credits | None | $5/month |
| Image/video generation | No | Yes |
| Labs access | No | Yes |
| Price | $0 | $0 for 12 months |

---

## Additional Tips

- **Verify from desktop** — the SheerID verification flow works more reliably in a desktop browser than on mobile
- **Check spam folder** — if your verification email does not arrive, check spam/junk folders
- **API for side projects** — the $5/month credit is enough to prototype a search-augmented AI feature using Sonar. Embed real-time web search with citations into your apps
- **Stack with referrals early** — refer classmates right after activating to maximize your free period
- **Re-verify before expiry** — if you are still enrolled, start the re-verification process before your 12 months expire to avoid any gap in access
- **Alternative paths to free Pro** — Perplexity also offers free Pro through T-Mobile and various partner promotions. Check the [Partner Promotions page](https://www.perplexity.ai/help-center/en/articles/11385821-partner-promotions-and-referral-programs) for current offers

---

**Sources:**
- [Perplexity Education Page](https://www.perplexity.ai/education)
- [What is Education Pro? — Perplexity Help Center](https://www.perplexity.ai/help-center/en/articles/12590157-what-is-education-pro)
- [How to Verify Your Student Status — Perplexity Help Center](https://www.perplexity.ai/help-center/en/articles/12162294-how-to-verify-your-student-status)
- [Student Referrals — Perplexity Help Center](https://www.perplexity.ai/help-center/en/articles/10964633-student-referrals)
- [SheerID and Perplexity Partnership — SheerID](https://www.sheerid.com/press-releases/sheerid-and-perplexity-partner-to-empower-academic-research/)
- [Perplexity API Documentation](https://docs.perplexity.ai/)
- [Perplexity Subscription Plans — Help Center](https://www.perplexity.ai/help-center/en/articles/11187416-which-perplexity-subscription-plan-is-right-for-you)
- [SheerID Verification for Perplexity](https://shop.sheerid.com/offers/free-perplexity-pro-for-students/)

## Submission 63: Paperspace Gradient Free Tier (by DigitalOcean)

- Category: Free GPU & Compute
- Source Link: https://www.paperspace.com/pricing
- Amount: Free GPU notebooks (M4000 8GB VRAM reliably; higher-end GPUs subject to availability); 5GB persistent storage; 6-hour sessions with unlimited restarts
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Paperspace Gradient (now part of DigitalOcean) offers **free GPU-powered Jupyter notebooks** for machine learning experimentation. The free tier includes access to NVIDIA GPUs — the reliably available **Quadro M4000 (8 GB VRAM)** plus nominally free Ampere GPUs (A4000, A5000, A6000) that are heavily constrained by availability. You get **5 GB of persistent storage**, 6-hour session limits, and unlimited session restarts. No weekly GPU hour cap. Credit card required for account verification.

> **Reality check:** While Paperspace advertises free access to high-end Ampere GPUs, in practice the M4000 is the GPU you will reliably get. Free A4000/A5000/A6000 instances exist but are frequently unavailable due to demand. The A100 is **not free** — it requires a paid plan plus $3.09/hr. Treat this as a free M4000 notebook platform with occasional access to better hardware.

---

## Registration (Step-by-Step)

1. Go to [paperspace.com](https://www.paperspace.com/) and click **Sign Up**
2. Create an account with email or sign in via Google/GitHub
3. Enter a valid **credit card** for verification (required even for the free tier; Paperspace uses Stripe and accepts Visa, MasterCard, Amex, Discover)
4. Navigate to **Gradient > Notebooks** from the dashboard
5. Click **Create Notebook**, choose a runtime template (e.g., PyTorch, TensorFlow, or Start from Scratch)
6. Under **Machine**, select one of the **Free** GPU/CPU options (Free-GPU M4000, Free-CPU C4, or check if Free-A4000/A5000/A6000 are available)
7. Click **Start Notebook** — your Jupyter environment launches in-browser

---

## Free Tier GPU Options

| Instance | GPU | VRAM | CPU Cores | RAM | Availability |
|---|---|---|---|---|---|
| **Free-GPU (M4000)** | NVIDIA Quadro M4000 | 8 GB | 8 | 8 GB | Reliably available |
| **Free-CPU (C4)** | None (CPU only) | — | 8 | 8 GB | Reliably available |
| **Free-P4000** | NVIDIA Quadro P4000 | 8 GB | 8 | 8 GB | Limited |
| **Free-P5000** | NVIDIA Quadro P5000 | 16 GB | 8 | 16 GB | Limited |
| **Free-RTX4000** | NVIDIA RTX4000 | 8 GB | 8 | 8 GB | Limited |
| **Free-RTX5000** | NVIDIA RTX5000 | 16 GB | 8 | 16 GB | Limited |
| **Free-A4000** | NVIDIA RTX A4000 | 16 GB | 8 | 45 GB | Rarely available |
| **Free-A5000** | NVIDIA RTX A5000 | 24 GB | 8 | 45 GB | Rarely available |
| **Free-A6000** | NVIDIA RTX A6000 | 48 GB | 8 | 45 GB | Rarely available |
| **Free-A100-80G** | NVIDIA A100 | 80 GB | 12 | 90 GB | Rarely available |

The M4000 uses the older Maxwell architecture. For reference, it benchmarks roughly 16x slower than an A100 on transformer workloads like BERT. The M4000 is comparable to or slightly slower than Google Colab's free T4.

---

## Limitations

- **6-hour auto-shutdown** — each session automatically terminates after 6 hours. You can restart immediately and run unlimited sessions
- **1 concurrent notebook** — free tier allows only one running notebook at a time
- **5 GB persistent storage** — persists between sessions. Overages cost $0.29/GB/month
- **Up to 5 projects** — free tier limits you to 5 Gradient projects
- **Public notebooks only** — free tier notebooks are publicly visible. Do **not** put API keys, credentials, or proprietary code in free notebooks
- **No terminal access** — free tier does not include notebook terminal/shell access
- **GPU availability not guaranteed** — free instances are shared. If demand is high, you may be queued or unable to start a session, especially for Ampere GPUs
- **No Deployments or Workflows** — free machines can only be used for Notebooks, not for Gradient Deployments or Workflows

---

## Paid Subscription Plans

| Feature | Free ($0/mo) | Pro ($8/mo) | Growth ($39/mo) |
|---|---|---|---|
| Storage | 5 GB | 15 GB | 50 GB |
| Projects | Public only | Private | Private |
| Auto-shutdown | 6 hours (fixed) | 12 hours (configurable) | 12 hours (configurable) |
| Free GPUs | M4000 + limited others | P4000, P5000, RTX4000, RTX5000 | All free-tier GPUs + priority |
| Concurrent notebooks | 1 | Multiple | Multiple |
| Terminal access | No | Yes | Yes |
| Support | Standard | Standard | Expert |

Paid on-demand GPU pricing (in addition to subscription):

| GPU | Hourly Rate |
|---|---|
| NVIDIA T4 | ~$0.45/hr |
| NVIDIA A4000 | ~$0.76/hr |
| NVIDIA A100 (40 GB) | $3.09/hr |
| NVIDIA A100 (80 GB) | $3.18/hr |
| NVIDIA H100 | $5.95/hr |

Multi-GPU configurations (2x, 4x, 8x) are available for A4000, A5000, A6000, and A100.

---

## Datacenter Regions

| Region | Code | Location |
|---|---|---|
| US-East | NY2 | New York, USA |
| US-West | CA1 | California, USA |
| Europe-West | AMS1 | Amsterdam, Netherlands |

Free GPU availability varies by region and time of day. A100 and H100 instances are concentrated in NY2.

---

## Comparison with Alternatives

| Feature | Paperspace Free | Google Colab Free | Kaggle Notebooks |
|---|---|---|---|
| GPU | M4000 (8 GB) | T4 (15 GB) | T4 (15 GB) or P100 |
| Session limit | 6 hours | ~12 hours (variable) | 12 hours |
| Weekly GPU limit | None | ~12 hours (soft) | 30 hours |
| Storage | 5 GB persistent | Limited (Google Drive) | 20 GB |
| CPU cores | 8 | 2 | 2 |
| Privacy | Public only | Private | Private |
| Idle timeout | None (runs full 6 hrs) | ~90 min idle disconnect | None |

**Key advantage:** No weekly GPU hour cap and no idle disconnects. **Key disadvantage:** Weaker GPU (M4000 vs T4) and notebooks are public.

---

## Additional Tips

- **Use environment variables carefully** — free notebooks are public, so never hardcode API keys. Use Gradient Secrets if available on your plan, or pass credentials at runtime
- **Save checkpoints to persistent storage** — your /storage directory (5 GB) persists between sessions. Save model checkpoints there to avoid losing work at the 6-hour shutdown
- **Check GPU availability before starting** — in the notebook creation screen, unavailable free GPUs will be grayed out. If the M4000 is also unavailable, try a different region or time of day
- **Pre-built templates save time** — Paperspace offers templates for PyTorch, TensorFlow, Hugging Face, Stable Diffusion, and fast.ai that come pre-configured with CUDA and dependencies
- **Consider Colab for better GPU** — if you need a T4 (stronger than M4000) and can tolerate idle disconnects, Google Colab's free tier may be a better fit for GPU-intensive work
- **Upgrade path** — the Pro plan at $8/month is reasonable if you need private notebooks and access to P4000/P5000/RTX4000 GPUs at no extra hourly cost
- **DigitalOcean integration** — since Paperspace was acquired by DigitalOcean, billing is handled through DigitalOcean. Existing DigitalOcean customers can access Paperspace through their DO account
- **Referral credits** — Paperspace occasionally offers referral credits ($10-$15) that can be applied toward paid GPU usage

---

**Sources:**
- [Paperspace Pricing](https://www.paperspace.com/pricing)
- [DigitalOcean Paperspace Pricing Documentation](https://docs.digitalocean.com/products/paperspace/pricing/)
- [Gradient Free GPU Benchmarks](https://blog.paperspace.com/gpu-benchmarking/)
- [Gradient vs Kaggle Comparison](https://blog.paperspace.com/gradient-kaggle-notebook-comparison/)
- [Gradient Community Notebook Guide](https://blog.paperspace.com/gradient-community-notebook-guide/)
- [Free Instances Documentation (GitHub)](https://github.com/Paperspace/Docs/blob/master/instances/instance-types/free-instances.md)
- [Paperspace Datacenter Regions](https://docs-next.paperspace.com/compute/regions)
- [Paperspace Free GPU Launch Announcement](https://blog.paperspace.com/paperspace-launches-gradient-community-notebooks/)

## Submission 62: Oracle Cloud Free Trial - $300 Credits

- Category: Cloud Providers
- Source Link: https://www.oracle.com/cloud/free
- Amount: $300
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

New Oracle Cloud Infrastructure (OCI) customers receive **$300 in free credits valid for 30 days** to explore all eligible OCI services, including Generative AI, Compute, Autonomous Database, and networking. After the trial, your account transitions to **Always Free** tier with 20+ services at no cost and no time limit. Credit card required for most users (identity verification only — you will not be charged unless you manually upgrade). **Accelerated onboarding** (no credit card) is available for students via Oracle Academy, Oracle CloudWorld attendees, and prospects engaged with Oracle Sales.

---

## Registration (Step-by-Step)

1. Go to [signup.oraclecloud.com](https://signup.oraclecloud.com/) or [oracle.com/cloud/free](https://www.oracle.com/cloud/free/)
2. Enter your **name, email address**, and **country**
3. Verify your email address via the confirmation link sent to your inbox
4. Set a **password** for your Oracle Cloud account
5. **Choose your home region** carefully — Always Free resources (Autonomous DB, Compute) can only be provisioned in your home region. This cannot be changed later
6. **Credit card verification** — enter a valid credit or debit card. Oracle will not charge your card unless you explicitly upgrade to a paid account
7. Review and accept Oracle Cloud Terms of Service
8. Account is provisioned — you receive $300 in credits immediately

**Accelerated onboarding (no credit card):**
- **Oracle Academy** students and educators
- **Oracle CloudWorld** event attendees
- Prospective customers engaged with **Oracle Sales**
- These users see a simplified signup flow with fewer verification steps

**Important:**
- One free trial per person
- The 30-day clock starts at signup, not at first usage
- Credits are in USD (or local currency equivalent in select countries)
- Choose your home region wisely — it determines where Always Free resources live

---

## What You Can Use the $300 Credits On

The credits work across **all eligible OCI services** during the 30-day window, including:

- **Compute** — VMs with AMD, Intel, or Arm (Ampere) processors, including GPU shapes
- **OCI Generative AI** — managed LLM service with Cohere, Meta Llama, and other models
- **Autonomous Database** — fully managed Oracle database with AI capabilities
- **Container Instances & Kubernetes (OKE)** — managed container orchestration
- **Block Storage, Object Storage, File Storage**
- **Networking** — VCN, Load Balancer, VPN, FastConnect
- **Data Science** — managed notebooks, model training, and deployment
- **AI Services** — Vision, Speech, Language, Document Understanding, Anomaly Detection

---

## Available AI & Generative AI Services

### OCI Generative AI Service (Managed LLMs)

A fully managed service for integrating large language models via API, SDK, CLI, or the Console playground. Available models include:

| Provider | Models | Capabilities |
|---|---|---|
| **Meta** | Llama 4 Maverick, Llama 4 Scout, Llama 3.3 (70B), Llama 3.2 90B Vision, Llama 3.2 11B Vision, Llama 3.1 (405B) | Chat, code, multimodal, multilingual, tool-calling |
| **Cohere** | Command A Reasoning, Command A Vision, Command A, Command R/R+ | Chat, summarization, RAG, enterprise search |
| **Google** | Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash-Lite | Multimodal, reasoning, code |
| **OpenAI** | gpt-oss-120b, gpt-oss-20b | General-purpose chat and reasoning |
| **xAI** | Grok 4, Grok 3 and variants | Chat, reasoning |

**Pricing model:** On-demand pay-per-character (prompt + response). Dedicated AI clusters available for hosting or fine-tuning (minimum 744 unit-hours commitment).

**Note:** Model availability varies by region. Check OCI documentation for your home region's supported models.

### Other OCI AI Services

| Service | Description |
|---|---|
| **OCI Vision** | Image classification, object detection, OCR |
| **OCI Speech** | Speech-to-text transcription |
| **OCI Language** | Sentiment analysis, entity recognition, translation |
| **OCI Document Understanding** | Extract text and structure from documents |
| **OCI Anomaly Detection** | Time-series anomaly detection |
| **OCI Data Science** | Managed JupyterLab notebooks, model training/deployment |
| **Agent Hub** | Build and deploy AI agents with RAG pipelines |

---

## Always Free Tier (No Time Limit)

After the 30-day trial, these services remain **free forever** with no credit card charges:

### Compute

| Resource | Always Free Limit |
|---|---|
| **AMD Compute VMs** | 2 VMs, each 1/8 OCPU + 1 GB RAM |
| **Arm Ampere A1 Compute** | 4 OCPUs + 24 GB RAM total (flexible: 1 large VM or up to 4 smaller VMs) |
| **Boot Volume Storage** | 200 GB total |
| **Block Volume Storage** | 200 GB total |

### Database

| Resource | Always Free Limit |
|---|---|
| **Autonomous Database** | 2 databases, 1 OCPU + 20 GB storage each |

### Networking & Other

| Resource | Always Free Limit |
|---|---|
| **Outbound Data Transfer** | 10 TB/month |
| **Object Storage** | 20 GB |
| **Load Balancer** | 1 instance (10 Mbps) |
| **VCN, Subnets, Security Lists** | Included |
| **Monitoring & Notifications** | 500M ingestion datapoints, 1M notification delivery |
| **Logging** | 10 GB/month |
| **Vault** | 20 key versions, up to 150 secrets |

The **Arm Ampere A1** offering (4 OCPUs + 24 GB RAM) is notably more generous than comparable free tiers from AWS, Azure, or Google Cloud.

---

## What Happens After the 30-Day Trial

1. **If you do nothing:** Your account stays active. Always Free resources continue running without interruption. Paid (non-free-tier) resources enter a **30-day grace period** — you can still use them but cannot create new paid resources. After the grace period, paid resources are **reclaimed** (deleted)
2. **If you upgrade to Pay As You Go:** You keep Always Free resources at no cost. You pay only for usage beyond free tier limits. No minimums, no prepayment. All service restrictions are lifted. Upgrade is **irreversible** (no downgrade option)
3. **Always Free services continue** regardless of whether you upgrade
4. **Support:** Free Tier accounts cannot open support tickets. Upgrading to Pay As You Go unlocks Oracle Support access

---

## Tips & Recommendations

- **Home region selection is permanent** — research which region has the services you need (especially for Generative AI model availability) before signing up
- **Arm Ampere A1 VMs are the best value** — 4 OCPUs and 24 GB RAM for free is exceptional. Use these for hosting side projects, dev environments, or lightweight production workloads
- **Set up budget alerts** if you upgrade to Pay As You Go — OCI has no automatic spending caps
- **Delete non-free resources** before your trial ends if you don't plan to upgrade, to avoid the grace period countdown
- **Oracle Academy** is the easiest path for students to bypass credit card verification — check if your institution participates
- **Trial extensions are possible** — contact Oracle Sales via the Upgrade page in the Console. Sales reps can extend trials or issue additional credits on a case-by-case basis
- **Always Free availability is capacity-limited** — popular regions may have limited Always Free Compute capacity. If you can't create an Always Free VM, try again later or consider a less popular region
- **OCI CLI and Terraform** are fully supported on Free Tier for infrastructure-as-code workflows
- **10 TB/month outbound transfer** is far more generous than other cloud providers' free tiers

---

**Sources:**
- [Oracle Cloud Free Tier](https://www.oracle.com/cloud/free/)
- [Oracle Cloud Free Tier FAQ](https://www.oracle.com/cloud/free/faq/)
- [OCI Free Tier Documentation](https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier.htm)
- [Always Free Resources](https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier_topic-Always_Free_Resources.htm)
- [What Happens When the Promotion Expires](https://docs.oracle.com/en-us/iaas/Content/GSG/Tasks/signingup_topic-What_Happens_When_the_Promotion_Expires.htm)
- [OCI Generative AI Service](https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/)
- [OCI Generative AI Pricing](https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing/)
- [OCI AI Services](https://www.oracle.com/artificial-intelligence/ai-services/)
- [Pretrained Foundational Models in OCI Generative AI](https://docs.oracle.com/en-us/iaas/Content/generative-ai/pretrained-models.htm)

## Submission 61: Oracle Cloud Always Free Tier

- Category: Cloud Providers
- Source Link: https://www.oracle.com/cloud/free
- Amount: Perpetual free compute, database, and storage resources (no dollar value; usage-limited). Additionally $300 in trial credits for 30 days.
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Oracle Cloud Infrastructure (OCI) offers a **perpetual Always Free Tier** with no time limit -- resources remain available indefinitely, even after the 30-day / $300 trial credits expire. The Always Free Tier includes **4 ARM Ampere A1 OCPUs with 24 GB RAM**, **2 AMD micro VMs**, **2 Autonomous AI Databases**, **200 GB block storage**, networking with **10 TB/month egress**, and 20+ additional services. No charges unless you explicitly upgrade and exceed free limits.

---

## Signup

1. Go to [oracle.com/cloud/free](https://www.oracle.com/cloud/free/)
2. Create an Oracle Cloud account with a valid email
3. Enter a credit/debit card for identity verification (no charge)
4. Select your **home region** carefully -- most Always Free resources can only be created in your home region, and it cannot be changed later
5. You receive $300 in trial credits (valid 30 days) plus immediate access to all Always Free services

**Important:** Choose a less popular home region (e.g., US Midwest (Chicago), Canada Southeast (Montreal), Germany Central (Frankfurt)) to avoid "Out of Capacity" errors when provisioning ARM instances. Popular regions like US East (Ashburn) and India often have exhausted free-tier ARM capacity.

---

## Compute

| Resource | Specification |
|---|---|
| **AMD Micro VMs** | 2x VM.Standard.E2.1.Micro -- 1/8 OCPU (burstable), 1 GB RAM each |
| **ARM Ampere A1** | VM.Standard.A1.Flex -- 4 OCPUs + 24 GB RAM total (3,000 OCPU-hours + 18,000 GB-hours/month) |
| **Max instances** | Up to 4 total (any combination of AMD + ARM) |
| **Boot volume** | 47 GB minimum per instance, 50 GB default |
| **AMD bandwidth** | 50 Mbps internet, 480 Mbps internal per VM |
| **ARM bandwidth** | Scales with OCPU count |
| **OS options** | Oracle Linux, Ubuntu, CentOS, Oracle Linux Cloud Developer |

### ARM Instance Flexibility

The A1.Flex shape lets you divide 4 OCPUs and 24 GB RAM across instances however you choose:

| Configuration | OCPUs | RAM | Use Case |
|---|---|---|---|
| 1 large VM | 4 | 24 GB | Self-hosted apps, databases |
| 2 medium VMs | 2 each | 12 GB each | Staging + production |
| 4 small VMs | 1 each | 6 GB each | Microservices, testing |

---

## Storage

| Resource | Limit |
|---|---|
| **Block Volume** | 200 GB total (boot + data volumes combined) |
| **Volume Backups** | 5 backups (boot + block combined) |
| **Object Storage** | 20 GB combined (Standard + Infrequent Access + Archive tiers) |
| **Object Storage API** | 50,000 requests/month |

---

## Database

| Resource | Limit |
|---|---|
| **Autonomous AI Database** | 2 instances (Transaction Processing, JSON, APEX, Analytics, or Mixed Workloads) |
| **CPU per DB** | 1 OCPU (non-scalable) |
| **Storage per DB** | 20 GB (non-scalable) |
| **Concurrent sessions** | 20 per instance |
| **NoSQL Database** | 3 tables, 25 GB/table, 133M reads + 133M writes/month |
| **MySQL HeatWave** | 1 standalone DB, 50 GB data storage, 50 GB backup storage |

### Autonomous Database AI Features

The Always Free Autonomous Database runs Oracle Database 23ai (upgradable to 26ai) and includes:

- **AI Vector Search** -- native VECTOR data type for storing and querying embeddings directly in the database
- **Hybrid Search** -- combine vector similarity search with relational, JSON, spatial, and graph queries in a single SQL statement
- **RAG support** -- built-in retrieval-augmented generation pipelines using your private business data
- **Vector indexes** -- approximate nearest neighbor (ANN) indexes for accelerated similarity search
- **Embedding generation** -- generate embeddings via ONNX models or REST APIs (Cohere, OpenAI, Google, Hugging Face, OCI Generative AI)
- **Select AI** -- natural language to SQL queries
- **Oracle APEX** -- low-code app builder included (2 ECPU, 20 GB, supports ~3-6 concurrent users)
- **Oracle Machine Learning** -- in-database ML algorithms
- **Graph Studio** -- property graph analytics
- **JSON Document Store** -- MongoDB-compatible API

---

## Networking

| Resource | Limit |
|---|---|
| **Virtual Cloud Networks (VCNs)** | 2 (IPv4 + IPv6) |
| **Outbound data transfer** | 10 TB/month |
| **Flexible Load Balancer** | 1 instance, 10 Mbps, 16 listeners, 1,024 backends |
| **Network Load Balancer** | 1 instance (unlimited bandwidth), 50 listeners, 1,024 backends |
| **Site-to-Site VPN** | 50 IPSec connections |
| **Reserved Public IPs** | 1 permanent |
| **Bastion Service** | Free (secure SSH access without jump hosts) |

The **10 TB/month outbound data** is significantly more generous than AWS (100 GB), GCP (200 GB), or Azure (100 GB) free tiers.

---

## Security & Identity

| Resource | Limit |
|---|---|
| **Certificates (CAs)** | 5 Certificate Authorities |
| **Certificates** | 150 certificates |
| **Vault (software keys)** | Unlimited software-protected master encryption keys |
| **Vault (HSM keys)** | 20 HSM-protected master encryption key versions |
| **Secrets** | 150 secrets |
| **Security Zones** | Free (enforce security best practices per compartment) |
| **Cloud Guard** | Free security posture management |
| **IAM** | Free (identity and access management, MFA) |

---

## Monitoring & Observability

| Resource | Limit |
|---|---|
| **Monitoring** | 500M ingestion + 1B retrieval data points |
| **Logging** | 10 GB/month |
| **APM (tracing)** | 1,000 events/hour |
| **Synthetic Monitoring** | 10 runs/hour |
| **Notifications (HTTPS)** | 1M/month |
| **Notifications (email)** | 1,000/month |
| **Email Delivery** | 3,000 emails/month |
| **Connector Hub** | 2 connectors |
| **Console Dashboards** | 100 dashboards |

---

## Additional Always Free Services

| Service | Limit |
|---|---|
| **Resource Manager (Terraform)** | 100 stacks, 2 concurrent jobs |
| **Fleet Application Management** | 25 resources/month |
| **OCI Cloud Shell** | Free browser-based terminal |
| **VCN Flow Logs** | Included in 10 GB/month logging |

---

## Idle Instance Reclamation Policy

Oracle **may reclaim idle Always Free compute instances**. An instance is considered idle if, over a 7-day period, ALL of the following are true:

| Metric | Idle Threshold (95th percentile) |
|---|---|
| CPU utilization | Below 10-20% |
| Network utilization | Below 10% |
| Memory utilization (A1 only) | Below 10% |

**What happens:**
1. You receive an email warning 7 days before reclamation
2. The instance is **stopped** (not deleted)
3. You can restart it if capacity is available in your region

**How to avoid reclamation:**
- **Upgrade to Pay As You Go (PAYG)** -- best option. You still get all Always Free resources at no cost, but instances are exempt from reclamation. No charges unless you exceed free limits
- **Right-size instances** -- don't allocate 4 OCPUs / 24 GB if your workload only needs 1 OCPU / 6 GB. Smaller instances reach utilization thresholds more easily
- **Run real workloads** -- web servers, DNS, VPN, databases, or CI/CD runners naturally generate enough utilization
- **Avoid artificial load generators** -- tools like `lookbusy` or `stress-ng` may violate Oracle's Terms of Service and risk account suspension

---

## Comparison with Other Cloud Free Tiers

| Feature | Oracle Cloud | AWS | GCP | Azure |
|---|---|---|---|---|
| **Free compute duration** | Perpetual | 12 months (t2.micro) | Perpetual (e2-micro) | 12 months (B1s) |
| **ARM instances** | 4 OCPU / 24 GB | None free | None free | None free |
| **Free databases** | 2 Autonomous DBs | None perpetual | None perpetual | None perpetual |
| **Monthly egress** | 10 TB | 100 GB | 200 GB | 100 GB |
| **Block storage** | 200 GB | 30 GB (12 mo) | 30 GB | 64 GB (12 mo) |

---

## Tips for Maximizing Value

- **Choose a less popular home region** to avoid ARM capacity issues. Check region availability at [cloud.oracle.com](https://cloud.oracle.com) before signing up
- **Upgrade to PAYG immediately** after signup to avoid idle reclamation, gain access to larger ARM capacity pools, and unlock Oracle Support -- you still pay nothing within Always Free limits
- **Use the ARM instances for self-hosting**: 4 OCPUs + 24 GB RAM is enough for Nextcloud, Gitea, Plausible Analytics, Uptime Kuma, n8n, or small Kubernetes clusters (k3s)
- **Autonomous Database for AI projects**: use AI Vector Search to build RAG applications, semantic search, or recommendation engines with the included 20 GB storage
- **The 10 TB egress is ideal** for hosting media, APIs, or CDN origin servers
- **Port 25 is blocked by default** -- request an exemption via service limits if you need outbound SMTP
- **No SLA on Always Free** -- do not use for mission-critical production workloads without upgrading
- **Oracle Cloud credits for startups** -- [Oracle for Startups](https://www.oracle.com/startup/) offers up to $30,000 in additional credits

---

**Sources:**
- [Oracle Cloud Free Tier](https://www.oracle.com/cloud/free/)
- [Always Free Resources Documentation](https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier_topic-Always_Free_Resources.htm)
- [Oracle Cloud Free Tier FAQ](https://www.oracle.com/cloud/free/faq/)
- [Oracle AI Vector Search](https://www.oracle.com/database/ai-vector-search/)
- [Oracle Autonomous AI Database](https://www.oracle.com/autonomous-database/)
- [Oracle AI Database 26ai](https://www.oracle.com/database/ai-native-database-26ai/)
- [OCI Free Tier Overview](https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier.htm)
- [Oracle for Startups](https://www.oracle.com/startup/)

## Submission 60: Oracle Academy Cloud Program - $300 Credits

- Category: Education & Student
- Source Link: https://academy.oracle.com/en/solutions-cloud-program.html
- Amount: $300 in cloud credits for 365 days + Always Free tier (unlimited duration)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Students and educators at Oracle Academy member institutions get **$300 in free Oracle Cloud credits valid for 365 days** plus permanent access to Always Free tier resources. Unlike the standard Oracle Cloud Free Trial (which gives the same $300 but expires in 30 days), the Academy program extends the credit window to a full year. **No credit card required** for Academy members -- Oracle recognizes them during signup and skips the card verification step. Always Free services continue indefinitely after credits expire.

The catch: your school must be an Oracle Academy Institutional member (free to join, but requires an accredited institution and faculty-initiated registration). Individual students cannot sign up on their own -- an educator at your institution must set it up first.

---

## Registration (Step-by-Step)

### Step 1 -- Institution Joins Oracle Academy (Faculty/Admin)

1. Go to [academy.oracle.com](https://academy.oracle.com/en/membership-join-oracle-academy.html)
2. Click **"Join Now"**
3. Sign in with an existing Oracle account or create a new one (free)
4. Search for your institution in the directory
5. Enter faculty information and acknowledge the membership agreement
6. Submit -- Oracle staff reviews the application (3-5 business days)
7. Once approved, the institution is an Oracle Academy member

**Requirements for the institution:**
- Must be accredited or registered with the country's department of education
- The person signing up needs authorization to sign legal agreements for the institution
- Non-fraudulent applications only -- Oracle verifies employment at the stated institution

### Step 2 -- Educator Provisions Student Accounts

1. Educator logs into the Oracle Academy Member Site
2. Navigates to the Cloud Program section
3. Provisions student cloud accounts directly -- students receive access without needing to provide a credit card
4. Students must be the **age of legal majority** in their country to access a cloud account

### Step 3 -- Student Accesses Oracle Cloud

1. Student receives account credentials from the educator
2. Signs in to the [Oracle Cloud Console](https://cloud.oracle.com)
3. $300 in credits are available immediately, valid for 365 days
4. Always Free resources are active from day one and persist indefinitely

**Important:** Check if your institution is already a member in the [Member Directory](https://academy.oracle.com/en/membership-member-institutions.html) before asking faculty to register.

---

## What the $300 Credits Cover

The credits can be spent on **any eligible OCI service**, including:

- Compute instances (beyond Always Free limits)
- Block and object storage (beyond Always Free limits)
- Oracle Cloud Marketplace products
- Big Data services
- Kubernetes (OKE)
- NoSQL Database (beyond Always Free limits)
- Machine learning notebooks
- Analytics and integration services
- OCI Generative AI service (Cohere and Meta models)

Credits expire when the 365 days are up or the $300 is fully consumed, whichever comes first. Paid resources provisioned with credits are **reclaimed** when credits expire unless you upgrade to a paid account.

---

## Always Free Services (No Expiration)

These resources persist forever, even after credits run out, and do not require a paid account:

### Compute

| Resource | Limit |
|---|---|
| **AMD Compute VMs** | 2 VMs, each 1/8 OCPU + 1 GB RAM |
| **Arm Ampere A1 Compute** | 4 cores + 24 GB RAM total (1 large VM or up to 4 smaller VMs) |
| **Block Volume Storage** | 100 GB total (boot + block volumes combined), 5 backups |
| **Object Storage** | 20 GB (Standard + Infrequent Access + Archive combined) |
| **Data Egress** | 10 TB/month across all services |

### Database

| Resource | Limit |
|---|---|
| **Autonomous Database** | 2 instances (ATP, ADW, JSON DB, or APEX), each 1 OCPU + 20 GB storage |
| **NoSQL Database** | 133M reads/month, 133M writes/month, 25 GB/table (up to 3 tables) |

### Networking

| Resource | Limit |
|---|---|
| **Virtual Cloud Networks** | 2 VCNs |
| **Flexible Load Balancer** | 1 instance, 10 Mbps bandwidth |
| **Network Load Balancer** | 1 instance |
| **Site-to-Site VPN** | Up to 50 IPSec connections |

### Other Always Free Services

| Service | Limit |
|---|---|
| **APEX Application Development** | Included with Autonomous DB |
| **OCI Vault** | 20 master encryption key versions (HSM), 150 secrets |
| **Monitoring** | 500M ingestion data points, 1B retrieval data points |
| **Notifications** | 1M per month (HTTPS) |
| **Logging** | 10 GB/month |
| **Resource Manager (Terraform)** | Unlimited |
| **Certificates** | 5 CAs, 150 certificates |

---

## Key Restrictions and Gotchas

1. **Institution-gated access** -- You cannot join as an individual student. Your institution must be an Oracle Academy member, and an educator must provision your account. If your school is not a member, ask a professor to start the registration process.

2. **Home region is permanent** -- When your account is created, you pick a home region. You **cannot change it later**. Always Free services only run in your home region. Choose carefully.

3. **ARM instance availability** -- Arm-based Ampere A1 instances are frequently unavailable in popular regions due to high demand. You may need to try different regions or retry at off-peak hours.

4. **No Oracle Support** -- Free tier users are not eligible for Oracle Support. No SLA applies.

5. **No commercial use** -- Oracle Academy resources are strictly for teaching, learning, and academic research. Commercial use is prohibited under the membership agreement.

6. **Autonomous DB limitations** -- Always Free Autonomous Databases do not support long-term backups, manual backups, or restore functionality. Storage and OCPU cannot be scaled.

7. **Account reclamation** -- If your Always Free account sits completely idle for an extended period, Oracle may reclaim (terminate) idle compute instances. Keep a minimal workload running.

8. **Port 25 blocked** -- Outbound SMTP (port 25) is blocked by default. You need to open a service limits request for an exemption.

9. **One account per person** -- Only one Oracle Cloud Free Tier account is permitted per person.

10. **Institutional survey required** -- The member institution must complete a survey within 12 months of joining and annually thereafter. If membership lapses, student access may be affected.

11. **Age requirement** -- Students must be the age of legal majority in their country (typically 18) to access a cloud account.

12. **No downgrade from paid** -- If you upgrade to a paid account, there is no way to downgrade back to free tier.

---

## Oracle Academy Curriculum (Bonus)

Oracle Academy membership also includes free access to teaching and learning resources:

- Database Design and Programming with SQL
- Programming with PL/SQL
- Java Fundamentals / Foundations / Programming
- Oracle APEX Application Development
- Cloud infrastructure and data management courses
- Oracle professional certification preparation through the Career Center

These are full courses with labs, not just documentation.

---

## How This Compares to Standard Oracle Cloud Free Tier

| Feature | Standard Free Tier | Oracle Academy Cloud Program |
|---|---|---|
| **Credits** | $300 | $300 |
| **Credit duration** | 30 days | 365 days |
| **Credit card required** | Yes | No |
| **Always Free services** | Yes | Yes (same) |
| **Eligibility** | Anyone | Academy member institutions only |
| **Commercial use** | Yes (after upgrade) | No (academic only) |

---

## Additional Tips

- **Check the member directory first** -- Before asking faculty to register, check if your institution is already listed at [academy.oracle.com/en/membership-member-institutions.html](https://academy.oracle.com/en/membership-member-institutions.html)
- **Pick the right home region** -- Choose a region close to you geographically. Once set, it cannot be changed, and Always Free resources only run there
- **Use APEX for web apps** -- Oracle APEX is a powerful low-code platform included free with Autonomous Database. It is one of the strongest offerings in the Always Free tier for building web applications
- **Ampere A1 is the real value** -- 4 ARM cores and 24 GB RAM for free is extremely generous. This is enough to run small production workloads, Docker containers, or development environments
- **Set up budget alerts** -- Even on free tier, configure budget alerts in the OCI Console to monitor credit consumption
- **Oracle Cloud Innovation Accelerator** -- For more advanced projects, selected teams can apply for up to $5,000 in credits valid for 1 year through Oracle's higher education program
- **Alternative if your school is not a member** -- The standard Oracle Cloud Free Tier gives the same $300 credits (but only for 30 days) and requires a credit card. Always Free services are identical

---

**Sources:**
- [Oracle Academy Cloud Program](https://academy.oracle.com/en/solutions-cloud-program.html)
- [Oracle Academy Cloud Overview](https://academy.oracle.com/en/solutions-cloud.html)
- [Oracle Academy Join Page](https://academy.oracle.com/en/membership-join-oracle-academy.html)
- [Oracle Academy Member Directory](https://academy.oracle.com/en/membership-member-institutions.html)
- [Oracle Academy Membership Benefits](https://academy.oracle.com/en/membership-benefits.html)
- [Oracle Cloud Free Tier](https://www.oracle.com/cloud/free/)
- [Oracle Cloud Free Tier FAQ](https://www.oracle.com/cloud/free/faq/)
- [Always Free Resources Documentation](https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier_topic-Always_Free_Resources.htm)
- [Always Free Resource Details](https://docs.oracle.com/en-us/iaas/Content/FreeTier/resourceref.htm)
- [OCI Free Tier Overview](https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier.htm)

## Submission 59: OpenAI Researcher Access Program - Up to $1,000

- Category: AI API Free Tiers
- Source Link: https://openai.com/form/researcher-access-program
- Amount: Up to $1,000 (some grants may exceed this for larger projects)
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

The OpenAI Researcher Access Program provides **up to $1,000 in API credits** for researchers studying responsible AI deployment, risk mitigation, and societal impacts. Credits are valid for **12 months** and can be used with any of OpenAI's publicly available API models — including GPT-4.1, GPT-5 family, o3, and more. Applications are reviewed **quarterly** (March, June, September, December). This is one of the few ways to get free OpenAI API access without paying, but it requires a genuine research affiliation and a solid proposal. Some recipients have reported grants exceeding the standard $1,000 amount.

---

## Registration (Step-by-Step)

1. Go to [openai.com/form/researcher-access-program](https://openai.com/form/researcher-access-program/) — this redirects to the application portal
2. You will be taken to **SurveyMonkey Apply** (openai.smapply.org) — create an account on that platform if you do not already have one
3. Fill out the application form with:
   - Your **research question** and a description of your planned research
   - **Planned use of OpenAI products** — how you intend to use API access to support your research
   - **Institutional affiliation** — your university, research lab, or nonprofit details
4. Submit the application — you can submit at any time during the year
5. Wait for the next quarterly review cycle (March, June, September, or December)
6. If accepted, expect **4-6 weeks** after the review decision for credits to appear in your account
7. Once credits are granted, ensure your OpenAI API account has a valid **payment method** registered (see Payment Method section below)

**Important:**
- You can check application status on the SurveyMonkey Apply platform, but do not expect updates before the next review window closes
- You will receive an email notification when your application is reviewed
- There is no limit on how many times you can apply — if rejected, you can resubmit with improvements

---

## Research Areas of Interest

OpenAI prioritizes proposals in the following areas, though they welcome suggestions for additional focus topics:

| Area | Example Questions |
|---|---|
| **Alignment** | What objective is a model pursuing? How can it be better aligned with human preferences via design or fine-tuning? |
| **Fairness & Representation** | How should fairness performance criteria be established for language models? How can models better support fairness goals in deployed contexts? |
| **Interdisciplinary Research** | How can AI development draw on philosophy, cognitive science, sociolinguistics, and other fields? |
| **Interpretability / Transparency** | How do models work mechanistically? Can we extract latent knowledge, identify concepts, or predict surprising behavior? |
| **Misuse Potential** | How can API systems be misused? What red-teaming approaches help developers think about responsible deployment? |
| **Robustness** | How robust are large generative models to natural perturbations like rephrasing or typos? |

The listed questions are illustrative — proposals addressing different questions within these areas are welcome.

---

## Available Models

Credits can be applied toward **any publicly available OpenAI API model**, including:

- **GPT-5.2** — flagship reasoning model
- **GPT-5.1**, GPT-5.1 Codex — advanced reasoning and coding
- **GPT-5**, GPT-5 mini, GPT-5 nano — general-purpose family
- **GPT-4.1**, GPT-4.1 mini, GPT-4.1 nano — non-reasoning models with 1M token context
- **o3**, o3-mini, o3-pro — reasoning models
- **o4-mini** — fast reasoning
- **GPT Image 1.5** / GPT Image 1 — image generation
- **GPT-4o mini TTS**, Transcribe models — audio
- **Text embeddings** and moderation models

Some models (like GPT-5 full, GPT-5.1) may require additional access approval via OpenAI's limited access process.

---

## Eligibility Details

Who qualifies:
- **Researchers affiliated with a university** (faculty, postdocs, PhD students, graduate researchers)
- **Research lab members** at recognized research organizations
- **Nonprofits** conducting research activities (not operational support — you must be doing actual research, not using credits for chatbots or internal tools)
- Must be based in a **country where the OpenAI API is available** (160+ countries — see [supported countries list](https://platform.openai.com/docs/supported-countries))

Who gets priority:
- **Early-stage researchers** with limited financial and institutional resources
- Researchers **without corporate sponsorship** or alternative funding sources
- Proposals that align closely with the research areas listed above

Who does NOT qualify:
- Companies or startups (see OpenAI's startup programs instead)
- Nonprofits using credits for operational purposes rather than research
- Individuals without any institutional affiliation

---

## Payment Method Requirement

Even though you receive free credits, you may need to register a payment method on your OpenAI API account:

- If your account is on the **Free/Explore tier**, you must add a valid payment card to upgrade your account and unlock higher rate limits
- OpenAI places a temporary **$5 hold** on the card for verification — this is released and you are not charged
- Your actual API usage is **deducted from the granted credits**, not your card
- Without upgrading, you may be stuck with very low rate limits that make research impractical

---

## Review Timeline & Grant Process

| Step | Timeline |
|---|---|
| Submit application | Any time (rolling submissions) |
| Review cycles | March, June, September, December |
| Decision notification | During or shortly after the review month |
| Credits deposited | 4-6 weeks after decision |
| Credit validity | 12 months from grant date |

**Example:** If you submit in January, your application is reviewed in the March cycle. You might hear back in late March or early April, and credits would land in your account by April-May.

---

## Credit Expiration & Reapplication

- Credits **expire after 12 months** and cannot be extended or renewed
- OpenAI **cannot reinstate expired credits** under any circumstances
- If credits expire before you finish your work, you can **submit a new application** through the portal
- In the new application, include details about the progress of your research to strengthen your case
- There is no explicit limit on how many times you can reapply

---

## Usage Policies

- Researchers must comply with OpenAI's **Usage Policies** and all applicable laws
- Review OpenAI's **sharing & publication policy** before applying — there may be guidelines on how you can publish findings
- The program supports responsible research; using credits for purposes that could harm others violates the terms
- Credits are for research use, not commercial products or services

---

## Additional Tips

- **Write a focused proposal** — clearly state your research question, methodology, and why OpenAI API access is specifically needed. Vague or overly broad proposals are less likely to be accepted
- **Demonstrate limited funding** — if you lack corporate sponsorship or institutional funding for API costs, make this clear in your application
- **Some grants exceed $1,000** — community reports indicate that some researchers have received up to $5,000, likely for larger or more impactful projects
- **Life Science Research Special Access Program** — OpenAI also runs a [separate program](https://help.openai.com/en/articles/11826767-life-science-research-special-access-program) for life science researchers with different terms
- **Startup alternative** — if you are a startup founder rather than a researcher, consider the [Ramp x OpenAI partnership](https://ramp.com/rewards/openai) which offers up to $2,500 in API credits
- **No free credits for general developers** — OpenAI no longer provides automatic free credits for new API accounts. You must prepay or qualify for a program like this one
- **Application portal moved** — the portal was previously at grants.openai.com and is now at openai.smapply.org. Both URLs work

---

**Sources:**
- [OpenAI Researcher Access Program (Application Portal)](https://openai.smapply.org/prog/openai_researcher_access_program/)
- [OpenAI Researcher Access Program (Official Page)](https://openai.com/form/researcher-access-program/)
- [Researcher Access Program FAQ](https://help.openai.com/en/articles/10139500-researcher-access-program-faq)
- [OpenAI API Models](https://platform.openai.com/docs/models)
- [OpenAI API Supported Countries](https://platform.openai.com/docs/supported-countries)
- [Life Science Research Special Access Program](https://help.openai.com/en/articles/11826767-life-science-research-special-access-program)
- [Ramp x OpenAI Credits](https://ramp.com/rewards/openai)
- [Community Discussion on Researcher Access](https://community.openai.com/t/researcher-access-program-application/960324)

## Submission 58: OpenAI for Startups - $2,500-$5,000 Credits

- Category: Startup Programs
- Source Link: https://openai.com/startups
- Amount: $2,500-$5,000 (varies by partner channel)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

OpenAI for Startups gives VC-backed founders **$2,500–$5,000 in API credits**, rate limit upgrades, 1:1 time with OpenAI solutions engineers, exclusive events, and a direct line to the OpenAI Startup team. The credit amount depends on which partner channel you use: **$2,500 via Ramp** or **$5,000 via Founder Stack** (historically through Microsoft for Startups Founders Hub). The core program resources — Build Hours, cookbooks, guides, and community — are **free for all founders**, but API credits require VC backing or enrollment through a partner platform.

---

## How to Get Credits (Step-by-Step)

There are two main paths to receiving OpenAI API credits. Choose the one that matches your situation.

### Path A: Through a Partner VC ($2,500–$5,000)

1. Confirm your startup is funded by one of OpenAI’s **partner VCs**: Sequoia, a16z, Thrive Capital, Kleiner Perkins, or Conviction Partners (more may be added over time)
2. Contact your VC’s portfolio support or operations team and ask about **OpenAI for Startups credits**
3. Your VC will provide a **referral code** or direct enrollment link
4. Go to [openai.com/startups](https://openai.com/startups/) and follow the partner enrollment flow
5. Provide your **OpenAI Organization ID** (found at [platform.openai.com/settings/organization](https://platform.openai.com/settings/organization)) and a short description of what you’re building
6. Credits typically appear in your OpenAI billing dashboard within a few days

### Path B: Through Ramp ($2,500)

1. Go to [ramp.com/rewards/openai](https://ramp.com/rewards/openai)
2. Sign up for a **Ramp corporate card** (no annual fee for startups; requires venture funding)
3. Enroll in Ramp’s **startup partners program**
4. Claim the OpenAI credit offer through the **partner benefits dashboard**
5. Credits activate within **24–48 hours** of verification

**Important:**
- You need an existing **OpenAI Platform account** with an Organization ID before applying through any path
- Credits are applied to your OpenAI billing account as a **credit grant** — they offset API usage charges
- Once credits are exhausted, billing switches to standard **pay-as-you-go** pricing automatically
- One credit grant per organization — you cannot stack multiple startup grants on the same org

---

## What You Get Beyond Credits

API credits are only one part of the program. All founders (even without VC backing) get access to:

| Benefit | Details |
|---|---|
| **Build Hours** | Live virtual sessions led by OpenAI engineers covering APIs, agentic tool-calling, image generation, voice agents, and more. Includes hands-on demos with code repos |
| **Cookbooks & GitHub Repos** | Open-source prompt templates, evaluation guides, and starter projects at [github.com/openai/build-hours](https://github.com/openai/build-hours) |
| **Founder Guides** | Use-case discovery, evals design, and shipping playbooks |
| **Case Studies** | Real examples from startups like Cursor, Harvey, Gamma, HeyGen, and others |
| **Events & Community** | AMAs, webinars, DevDay sessions, meetups, and the monthly Startup Drop newsletter |

VC-backed startups in the partner network get **additional perks**:
- **Rate limit upgrades** — higher RPM/TPM limits on API models
- **1:1 sessions** with OpenAI solutions engineers
- **Exclusive events** and direct contact with the Startup team

---

## Available API Models

Credits can be used across **all OpenAI API models**. Key models as of early 2026:

| Model | Best For | Approx. Cost (per 1M tokens) |
|---|---|---|
| **GPT-5.2** | Flagship reasoning, complex tasks | Higher tier |
| **GPT-5 / GPT-5 mini** | Advanced reasoning | ~$40 / ~$8 |
| **o3** | Complex reasoning tasks | $2 input, $8 output |
| **o4-mini** | Fast, cost-efficient reasoning | Budget reasoning |
| **GPT-4.1** | Best non-reasoning model | $2 input |
| **GPT-4.1 mini / nano** | Cost-efficient general use | Sub-$1 input |
| **GPT-4o** | Legacy multimodal | $2.50 input, $10 output |
| **GPT-4o mini** | Ultra-cheap simple tasks | $0.15 input, $0.60 output |

**Tip:** $2,500 in credits goes a long way with GPT-4o mini (~4 billion input tokens) but burns fast with GPT-5.2. Plan your model usage carefully.

---

## Credit Expiration & Billing

- **Standard validity:** Credits expire **12 months** after issuance (per OpenAI’s [Service Credit Terms](https://openai.com/policies/service-credit-terms/))
- **No extensions:** OpenAI does not extend credit grant expiration dates under any circumstances
- **Non-transferable:** Credits cannot be transferred, sold, or converted to cash
- **After exhaustion:** Your account automatically switches to pay-as-you-go billing. If you don’t have a payment method on file, API calls will fail
- **No auto-charge surprise:** You won’t be charged unless you’ve added a payment method and have usage beyond the credit balance

---

## Partner VC Network

As of early 2026, the confirmed partner VCs are:

- **Sequoia Capital**
- **Andreessen Horowitz (a16z)**
- **Thrive Capital**
- **Kleiner Perkins**
- **Conviction Partners**

The list may expand over time. If your VC is not listed, ask them directly — some firms join the network without public announcements. You can also check with your VC’s portfolio services team.

---

## Microsoft for Startups (Historical Note)

Microsoft for Startups Founders Hub **previously** offered $2,500–$5,000 in direct OpenAI credits (at Level 4 membership). **This benefit has been discontinued.** Startups in the Microsoft program can still use their Azure Sponsorship credits with **Azure OpenAI Service**, which provides access to the same models at Azure’s enterprise pricing (often 20–30% cheaper than OpenAI direct pricing).

---

## Related OpenAI Programs

| Program | What It Is | Who It’s For |
|---|---|---|
| **OpenAI Grove** | Pre-idea program: dense talent network, co-building with OpenAI researchers, credits and VC intros | Technical talent at the very start of company building |
| **OpenAI Pioneers Program** | Custom fine-tuned models + domain-specific evals, working directly with OpenAI’s team | Established companies with high-value applied AI use cases |
| **OpenAI Startup Fund** | $100M venture fund investing in AI companies | Startups seeking investment (not credits) |
| **Researcher Access Program** | Up to $1,000 in API credits for 12 months | Researchers at academic institutions or nonprofits |

---

## Additional Tips

- **Ramp is the easiest path** if you have venture funding but your VC is not in OpenAI’s partner network. Sign up for the Ramp card, claim credits, done
- **Combine with Azure credits** — if you also qualify for Microsoft for Startups, use Azure OpenAI Service with Azure credits for production workloads and save your direct OpenAI credits for features only available on the OpenAI platform (e.g., Assistants API, real-time API)
- **Use the Batch API** to stretch credits further — it cuts costs by up to 50% for non-time-sensitive workloads
- **Set usage limits** in the OpenAI dashboard at [platform.openai.com/settings/organization/limits](https://platform.openai.com/settings/organization/limits) to avoid burning through credits accidentally
- **Build Hours recordings** are available on-demand even if you miss the live sessions
- **OpenAI Organization ID** is required for most credit applications — have it ready before you apply
- **Hackathon credits** — OpenAI sometimes distributes credits at hackathons and DevDay events. Attend these if you want additional grants

---

**Sources:**
- [OpenAI for Startups](https://openai.com/startups/)
- [Ramp x OpenAI Credits](https://ramp.com/rewards/openai)
- [OpenAI Service Credit Terms](https://openai.com/policies/service-credit-terms/)
- [OpenAI API Pricing](https://openai.com/api/pricing/)
- [OpenAI API Models](https://platform.openai.com/docs/models)
- [OpenAI Build Hours GitHub](https://github.com/openai/build-hours)
- [OpenAI Pioneers Program](https://openai.com/index/openai-pioneers-program/)
- [OpenAI Grove](https://openai.com/index/openai-grove/)
- [OpenAI Startup Fund](https://www.openai.fund/)
- [Microsoft for Startups — OpenAI Credits Discontinuation](https://learn.microsoft.com/en-us/answers/questions/2275712/is-openai-credits-available-as-part-of-startup-ben)

## Submission 57: NVIDIA Inception Program - Up to $100K+ Credits

- Category: Startup Programs
- Source Link: https://www.nvidia.com/en-us/startups
- Amount: Up to $100,000 (AWS) + up to $150,000 (Nebius) + $10,000 DLI training + 30% off DGX Cloud + GPU discounts
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

NVIDIA Inception is a **free virtual accelerator program** for AI startups, with no equity, fees, or cohort deadlines. It provides access to **up to $100,000 in AWS cloud credits** (via AWS Activate), **up to $150,000 in Nebius cloud credits**, **30% off DGX Cloud**, **$10,000 in Deep Learning Institute training credits**, preferred pricing on NVIDIA GPUs (up to 30% off), and VC introductions via Inception Capital Connect. Over 19,000 startups have been accepted globally, including notable alumni like OpenAI and Hugging Face.

---

## Registration (Step-by-Step)

1. Go to [nvidia.com/en-us/startups](https://www.nvidia.com/en-us/startups/)
2. Click **"Join Inception"** or **"Apply Now"**
3. Fill out the application form (~30 minutes). You will need:
   - Company name, website URL, and incorporation date
   - Description of your AI/ML product or solution
   - Funding stage and team size
   - Key competitors and company mission
4. **Upload a pitch deck** (PDF) covering your mission, product, team, and traction. NVIDIA uses this to determine how to best support your startup
5. Submit the application — it must be in **English**
6. Wait for NVIDIA to review your application (no publicly stated timeline, but most legitimate AI startups are accepted; allow approximately **2 weeks**)
7. Once approved, access your **personalized Inception portal** to request benefits like cloud credits and DLI training codes

**Important:**
- No application fees, membership fees, or equity requirements
- Rolling admissions — no deadlines or cohort structure
- You must **update your profile every 6 months** to remain eligible for benefits
- Contact: inceptionprogram@nvidia.com (for China: inception_cn@nvidia.com)

---

## Who Cannot Apply

The following types of organizations are **not eligible** for NVIDIA Inception:

- **Cryptocurrency-focused** companies (explicitly excluded since April 2025)
- **Consulting and outsourced development** firms
- **Cloud service providers**
- **Resellers and distributors**
- **Public companies**
- Companies **older than 10 years**

---

## Cloud Credits (Partner Programs)

Once accepted into Inception, you can request cloud credits through the **benefits request tab** in your Inception portal. Credits are provided through partner programs, not directly by NVIDIA.

| Partner | Credit Amount | Notes |
|---|---|---|
| **AWS Activate** | Up to $100,000 | Access NVIDIA GPUs on Amazon EC2. Must use same email for Inception and AWS accounts. Some startups report inconsistent approval |
| **Nebius AI Lift** | Up to $150,000 | Announced at GTC 2025. Includes $10K inference credits, early access to Blackwell GPUs. First 3 months of membership get extra perks |
| **FluidStack** | Up to $100,000 | CPU credits for training and serving frontier models |
| **Scaleway** | Varies | EU-focused cloud. Benefits for new Scaleway Startup Program members only |
| **Microsoft Azure** | Up to $150,000 (tiered) | Via Microsoft for Startups Founders Hub. Credits granted in tiers ($1K, $5K, $25K, etc.) as usage grows. Higher tiers require demo video, proof of funding, and 50%+ credit utilization |

**Caveat:** AWS credits are the most commonly advertised benefit, but many startups report difficulties obtaining the full $100K. AWS may reject applications even after NVIDIA Inception approval. The $100K tier may informally require $250K+ in institutional funding, though this is inconsistently applied.

---

## DGX Cloud Discount

Inception members get **30% off** the standard price of NVIDIA DGX Cloud.

- **Minimum commitment:** 4 nodes for 1 month
- **Minimum spend:** $75,000
- DGX Cloud is available through AWS, GCP, Azure, FluidStack, and other GPU cloud providers
- This benefit is primarily relevant for **later-stage startups** with significant compute budgets

### DGX Cloud Innovation Lab (for early-stage startups)

Select Inception members can apply to the **DGX Cloud Innovation Lab**, an exclusive 60-day accelerator offering:
- Hands-on access to NVIDIA DGX Cloud GPUs
- Integrated AI software and expert support
- Self-serve developer platform for rapid prototyping

---

## Deep Learning Institute (DLI) Training

Inception members receive **$10,000 in free DLI credits** plus discounts on expert-led workshops.

| Benefit | Details |
|---|---|
| **Free self-paced courses** | Redeemable codes provided via Inception portal |
| **Course topics** | AI/deep learning, data science, accelerated computing, robotics, graphics |
| **Course formats** | 2-hour technique courses, 8-hour project-based courses with certificate |
| **Instructor-led workshops** | Available at discounted pricing |
| **Certifications** | Complete assessments to earn NVIDIA DLI certificates |
| **GPU lab access** | Cloud-based GPU-accelerated servers included with courses |

Credits can be redeemed for self-paced courses, public/private workshops, Academy memberships, and certification exams.

---

## Hardware Discounts

Inception members get **preferred pricing (up to 30% off)** on select NVIDIA enterprise-grade GPUs.

- Available GPUs include: **DGX systems, A30, A40, A100 (Ampere), H100 (Hopper), RTX A6000**
- NVIDIA does **not sell directly** to Inception members — the discount functions as a **rebate** deducted from the standard price that any authorized resale partner provides
- Some members get **early access to new hardware releases**
- You must work with NVIDIA's authorized resale partners to purchase

---

## Inception Capital Connect & VC Network

Inception Capital Connect links eligible startups with NVIDIA's **VC Alliance** — a network of venture firms, private equity, family offices, and sovereign wealth funds.

- **How it works:** VCs browse a portal of Inception startups actively raising funds. With a single click, both parties are directly connected to discuss investment
- **Founding VC members:** NEA, Acrew, Mayfield, Madrona Venture Group, In-Q-Tel, Pitango, Vanedge Capital, OurCrowd
- **European VCs:** Accel, Elaia, Partech, Sofinnova Partners (eligible startups can receive up to $100K in GPU capacity credits)
- **Events:** Curated networking events bringing together startups, top VCs, and NVIDIA executives
- **Eligibility-based:** Not all members automatically get VC introductions. NVIDIA uses funding stage and startup maturity to determine fit

---

## Membership Tiers

| Tier | Requirements | Extra Benefits |
|---|---|---|
| **Community** | Default tier on acceptance | All standard benefits: cloud credits, DLI training, GPU discounts, co-marketing |
| **Premier** | Closed at least **Series A funding** + product **in production** + uses NVIDIA technology + passes review | Dedicated NVIDIA relationship manager, increased marketing support, Premier-only events, architectural guidance |

---

## Co-Marketing & Branding

- Official NVIDIA Inception **program badges** for your website and materials
- **Co-branded social content** and customizable assets for events
- Featured in NVIDIA **marketing campaigns, newsletters, and case studies**
- Access to NVIDIA's global networks and go-to-market opportunities
- Monthly newsletter with product updates and members-only offers

---

## Additional Tips

- **Application difficulty is low** — the program has a relatively inclusive acceptance process compared to selective accelerators. If your startup genuinely uses AI/ML, you will likely be accepted
- **Highlight AI/ML centrality** in your pitch deck — NVIDIA wants to see that AI is core to your product, not a peripheral feature
- **AWS credits are not guaranteed** — apply early and use the same email across NVIDIA Inception and your AWS account. If rejected, try again or escalate through the NVIDIA Developer Forums
- **Profile updates are mandatory** — if you don't update your Inception profile every 6 months, you lose access to benefits
- **No term limits** — once accepted, you remain a member indefinitely (as long as you maintain eligibility and update your profile)
- **Revenue not required** — pre-revenue startups at any funding stage are welcome
- **DGX Cloud discount is expensive** — the $75K minimum spend makes it impractical for early-stage startups. Look into the DGX Cloud Innovation Lab instead
- **Stack cloud credits** — you can potentially combine AWS Activate, Nebius AI Lift, FluidStack, Scaleway, and Azure credits from multiple partner programs
- **NVIDIA Connect (for ISVs)** — if you are an independent software vendor rather than a startup, check [NVIDIA Connect](https://www.nvidia.com/en-us/programs/isv/) instead

---

**Sources:**
- [NVIDIA Inception Program](https://www.nvidia.com/en-us/startups/)
- [NVIDIA Inception Guide (ThunderCompute)](https://www.thundercompute.com/blog/nvidia-inception-program-guide)
- [AWS & NVIDIA Startup Collaboration](https://aws.amazon.com/blogs/startups/accelerating-startup-growth-how-nvidia-and-aws-are-collaborating-to-grow-ai-startups/)
- [NVIDIA Inception VC Alliance](https://www.nvidia.com/en-us/startups/venture-capital/)
- [Nebius AI Lift for Inception Members](https://nebius.com/blog/posts/ai-lift-startups-innovation-with-nvidia)
- [FluidStack x NVIDIA Inception](https://www.fluidstack.io/nvidia_inception_offering)
- [Scaleway Startup Program for Inception](https://www.scaleway.com/en/scaleway-startup-program-for-nvidia-inception-members/)
- [NVIDIA DGX Cloud Innovation Lab](https://www.nvidia.com/en-us/data-center/dgx-cloud/innovation-lab/)
- [NVIDIA Deep Learning Institute](https://www.nvidia.com/en-us/training/)
- [NVIDIA Inception Application](https://programs.nvidia.com/phoenix/application)
- [Crypto Exclusion (CoinTelegraph)](https://cointelegraph.com/news/nvidia-excludes-crypto-startups-accelerator-program)
- [Cirrascale — NVIDIA Inception Overview](https://www.cirrascale.com/products-and-services/nvidia-inception)

## Submission 56: NVIDIA Graduate Fellowship - $60,000

- Category: Education & Student
- Source Link: https://research.nvidia.com/graduate-fellowships
- Amount: Up to $60,000 per academic year
- Difficulty: 5/5 (Very Hard)
- Expires At: N/A

### Description

The NVIDIA Graduate Fellowship Program awards **up to $60,000** to PhD students conducting research relevant to accelerated computing and NVIDIA technologies. Now in its 25th year, the program has awarded over 200 grants worth more than $7.3 million since 2002. The fellowship includes a **mandatory paid summer internship** at an NVIDIA research office before the fellowship year begins. Only about **10 students are selected** each cycle from a global applicant pool, making this one of the most competitive and generous fellowships in AI/GPU research. The award covers stipend, tuition, and mandatory fees and is disbursed through your university.

---

## Eligibility Requirements

- **Degree level:** Must be enrolled in a PhD program (not master's)
- **Progress:** Must have completed at least your **first year** of PhD studies at time of application
- **Fields:** Computer Science, Computer Engineering, System Architecture, Electrical Engineering, or a closely related area
- **Research status:** Must be actively working on your PhD thesis research
- **Enrollment:** Must be a **full-time PhD student** for the entire fellowship academic year (9 months) — you cannot be graduating before May/June of the fellowship year's end
- **Internship:** Must be available for an **in-person summer internship** (3-4 months) at one of NVIDIA's research offices
- **Exclusions:** Full-time NVIDIA employees and their immediate family members cannot participate
- **Geography:** Open to applicants **worldwide** — there is no citizenship or residency restriction

---

## Application Process (Step-by-Step)

1. Visit the official fellowship page at [research.nvidia.com/graduate-fellowships](https://research.nvidia.com/graduate-fellowships)
2. Wait for the application window to open (typically **early August** each year)
3. Prepare your **research summary / thesis proposal** — up to 2 pages plus bibliography (bibliography does not count toward the page limit)
4. Prepare your **CV/resume** including full contact information
5. Secure **2-3 professor nomination letters** — at least one must be from your thesis advisor, maximum 3 letters allowed
6. Confirm your **availability for the mandatory summer internship** at an NVIDIA research office
7. Submit everything through NVIDIA's online submission portal before the deadline
8. Recommenders must submit their letters by the separate recommender deadline (typically a few hours before the applicant deadline on the same day)
9. Wait for results — decisions are typically communicated by **end of November**
10. If selected, complete the **summer internship** before the fellowship year begins

**Key dates (2026-2027 cycle, for reference):**
- Application window: August 2025
- Application deadline: September 15, 2025, 3:00 PM Pacific
- Recommender deadline: September 15, 2025, 12:00 PM Pacific
- Notifications: End of November 2025
- Internship: Summer 2026
- Fellowship year: 2026-2027 academic year

**The 2026-2027 cycle is now closed.** Watch the official page starting in summer 2026 for the 2027-2028 application cycle, which will likely follow the same August-September timeline.

---

## Research Areas of Interest

NVIDIA looks for research aligned with accelerated computing and GPU innovation. Recent recipients have worked in:

| Area | Example Topics |
|---|---|
| **Deep Learning / AI** | Novel architectures, training efficiency, model security |
| **Robotics** | Generalist robots, sim-to-real transfer, manipulation |
| **Autonomous Systems** | Self-driving vehicles, drone navigation |
| **Computer Architecture** | GPU design, hardware accelerators, memory systems |
| **Computer Graphics** | Real-time rendering, neural radiance fields, simulation |
| **Programming Systems** | Compilers, DSLs for GPUs, performance optimization |
| **Security** | AI agent security, prompt injection defense, adversarial ML |
| **Healthcare / Science** | Drug discovery, medical imaging, computational biology |

---

## Selection Process

The review is conducted by a committee of **over 80 senior NVIDIA engineers and managers**, led by Chief Scientist **Bill Dally**. The review process takes approximately 8 weeks.

**Evaluation criteria:**
- **Student quality** — recommendations, GPA, academic track record
- **Research quality** — results to date, publications, proposal merit
- **NVIDIA relevance** — how closely your work connects to GPU technology and NVIDIA's research domains

**Competitive tips:**
- Having at least one strong publication in a top-tier venue significantly strengthens your application
- Explain not just your current research but your **long-term vision** — what you plan to do after graduation
- If you have ideas for your internship project at NVIDIA, mention them in your proposal
- Demonstrating familiarity with NVIDIA's technology stack (CUDA, cuDNN, TensorRT, etc.) helps

---

## Mandatory Summer Internship

The fellowship **requires completion of an in-person summer internship** at an NVIDIA research office before the fellowship year begins. This is not optional.

**Internship locations:**
- United States
- Canada
- Germany
- Sweden
- Finland
- Switzerland
- Israel
- Taiwan

**What to expect:**
- Duration: approximately **3-4 months** during summer
- You will work with NVIDIA researchers on projects related to your area of expertise
- The internship is paid separately from the fellowship award
- NVIDIA provides a **housing stipend** and covers travel costs if you are assigned to an office 50+ miles from your university
- Work-life balance is reasonable — expect roughly 40-hour weeks
- Access to NVIDIA hardware, proprietary tools, and mentorship from senior researchers

---

## Award Details

| Detail | Info |
|---|---|
| **Amount** | Up to $60,000 |
| **Duration** | One academic year (9 months) |
| **Coverage** | Stipend, tuition, and mandatory fees |
| **Disbursement** | Paid to the university, not directly to the student |
| **Overhead** | No overhead or indirect costs may be charged to the award |
| **Transferability** | Not transferable to another student |
| **Additional perks** | Access to NVIDIA products, technology, and mentorship |

The actual award amount is based on your individual costs for stipend, tuition, and mandatory fees — $60,000 is the maximum.

---

## 2026-2027 Recipients (for reference)

The most recent cohort of 10 fellows came from:
- Stanford University (4 recipients)
- University of Southern California
- UC San Diego
- UC Berkeley
- Harvard University
- Georgia Institute of Technology
- University of Washington

Four additional finalists were recognized from MIT, Peking University, Max Planck Institute, and Technical University of Darmstadt.

---

## Related NVIDIA Education Programs

If you don't qualify for the fellowship or want additional resources:

- **NVIDIA Developer Program** — free access to SDKs, tools, and one complimentary DLI self-paced course (worth up to $90). Open to everyone at [developer.nvidia.com/developer-program](https://developer.nvidia.com/developer-program)
- **DLI Teaching Kit Program** — free curriculum materials and DLI course codes for university educators and their students
- **NVIDIA Student Network** — community program with exclusive training, tools, and projects for AI students (requires faculty advisor and access to NVIDIA hardware)
- **Educational Hardware Pricing** — academic discounts on H100/H200 GPUs and Jetson developer kits
- **Jetson AI Certification** — free, open-source courses with certificates for NVIDIA Jetson and AI skills
- **Open Hackathons** — free events to accelerate research applications with NVIDIA mentors

---

## Additional Tips

- **Timeline awareness** — applications typically open in early August with a mid-September deadline. Set a calendar reminder for July to start preparing your materials
- **This is not a grant for equipment** — the award covers living/tuition costs so you can focus on research. If you need GPU hardware, look into NVIDIA's Academic Hardware Grant program separately
- **International students welcome** — the program is explicitly open worldwide, but you must be able to obtain authorization to work at an NVIDIA office for the summer internship (visa sponsorship details should be confirmed with NVIDIA)
- **Reapplication** — if you are not selected, you can apply again in a subsequent cycle as long as you still meet the eligibility requirements
- **Not for master's students** — this is strictly a PhD program. Master's students should look at the NVIDIA Student Network or DLI programs instead
- **Start early on recommendation letters** — you need 2-3 letters with a tight deadline. Give your recommenders at least 4-6 weeks notice

---

**Sources:**
- [NVIDIA Graduate Fellowship Program (Official)](https://research.nvidia.com/graduate-fellowships)
- [Applications Open for 2026-2027 Fellowship (NVIDIA Blog)](https://blogs.nvidia.com/blog/applications-open-graduate-fellowship-awards-2025/)
- [2026-2027 Fellowship Recipients Announced (NVIDIA Blog)](https://blogs.nvidia.com/blog/graduate-fellowship-recipients-2026-2027/)
- [NVIDIA Higher Education and Research Resources](https://developer.nvidia.com/higher-education-and-research)
- [NVIDIA Developer Program](https://developer.nvidia.com/developer-program)
- [NVIDIA Student Network](https://developer.nvidia.com/student-network)

## Submission 55: NVIDIA Academic Grant Program - 30,000 H100 GPU Hours

- Category: Education & Student
- Source Link: https://www.nvidia.com/en-us/industries/higher-education-research/academic-grant-program
- Amount: Up to 30,000 H100 80GB GPU hours + optional physical hardware (up to 8 RTX PRO 6000 GPUs or 2 DGX Spark supercomputers depending on research area)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

NVIDIA's Academic Grant Program provides **free cloud GPU hours, physical GPU hardware, and software grants** to academic researchers at accredited institutions. Depending on your research area, you can receive **up to 30,000 NVIDIA H100 80GB GPU hours**, up to **8 RTX PRO 6000 GPUs**, or up to **2 DGX Spark supercomputers** — all at no cost. This is a **competitive, proposal-based program** with quarterly submission windows. NVIDIA retains no IP rights over your work.

---

## Eligibility

- **You must hold a full-time faculty position** at an accredited institution that awards research PhDs. Postdocs, students, and adjuncts are **not eligible** to apply directly.
- Students and postdocs **can participate** on a funded project, but the PI (principal investigator) must be the applicant.
- Your project must incorporate **pretrained models from [ai.nvidia.com](https://ai.nvidia.com)** and/or make extensive use of **NVIDIA software distributions** (CUDA, cuDNN, RAPIDS, NeMo, Modulus, etc.).
- Submissions accepted **worldwide** from any accredited institution.
- Previous grant winners **must submit results** through the program portal before reapplying.
- Only **one submission per person per quarter** is allowed (maximum four per year).
- Each applicant can receive only **one award per calendar year**.

---

## Research Focus Areas & Resources

The program runs separate Calls for Proposals (CFPs) organized into focus areas. Each area offers different resource packages:

| Focus Area | Cloud GPU Hours | Physical Hardware |
|---|---|---|
| **Simulation & Modeling** (scientific simulation, quantum computing, physics-informed ML) | Up to 30,000 H100 80GB hours | Up to 8 RTX PRO 6000 GPUs |
| **Data Science** (data processing, analytics, operational research, graph analytics) | Up to 30,000 H100 80GB hours | Up to 2 DGX Spark supercomputers |
| **Robotics & Edge AI** (robotics, autonomous vehicles, 5G/6G, smart spaces, federated learning) | Up to 30,000 H100 80GB hours | Up to 2 AGX Developer Kits + up to 4 RTX PRO 6000 GPUs |

Additional CFPs may cover **Generative AI Training & Model Development**, **Generative AI Alignment & Inferencing**, and other emerging topics. Check the program page for currently active CFPs.

---

## Application Process (Step-by-Step)

1. Go to the [NVIDIA Academic Grant Program page](https://www.nvidia.com/en-us/industries/higher-education-research/academic-grant-program/) and review the currently active **Calls for Proposals (CFPs)**
2. Download and read the **CFP document** for your research area — each CFP has specific requirements
3. Download the **proposal template** — submissions that do not follow the template will be rejected without review
4. Write your proposal addressing: research objectives, methodology, how NVIDIA technology (models from ai.nvidia.com, NVIDIA software) will be used, timeline, and expected outcomes
5. Go to the [Academic Grants Portal](https://academicgrants.nvidia.com/academicgrantprogram/s/Application) and create an account or log in
6. Submit your proposal during an active submission window (see timeline below)
7. Wait for the decision — NVIDIA reviews proposals and announces results quarterly

**Important:**
- Applications are reviewed **immediately upon submission** — incomplete or non-conforming applications are declined on the spot
- Projects that do not align with the selected CFP will be declined
- Your project timeline must align with the expected GPU hours availability dates
- NVIDIA cannot provide individual feedback on declined proposals due to volume
- Watch the [Insider Tips webinar](https://www.nvidia.com/en-us/on-demand/session/other25-nvidia-agp/) on NVIDIA On-Demand for advice from program reviewers

---

## Submission Windows & Decision Timeline

| Submission Window | Decision Month | GPU Hours Available |
|---|---|---|
| January 1 – March 31 | June | July – December 31 |
| April 1 – June 30 | September | October – March 31 |
| July 1 – September 30 | December | January – June 30 |
| October 1 – December 31 | March | April – September 30 |

GPU hours are available for approximately **6 months** after the award decision. Hardware grants ship according to schedules described in the FAQ.

---

## How GPU Hours Are Delivered

Cloud GPU hours are delivered through **Saturn Cloud**, a managed ML platform:

1. After your grant is approved, you receive **login credentials via email** for the Saturn Cloud environment
2. Log in at the provided URL and access JupyterLab or connect via SSH using PyCharm/VSCode
3. Create a Python server resource and select your GPU configuration

### Available Machine Types (Saturn Cloud)

| Configuration | vCPUs | GPUs | Memory |
|---|---|---|---|
| 1x A100 | 16 | 1 | 256 GB |
| 2x A100 | 32 | 2 | 512 GB |
| 4x A100 | 64 | 4 | 1,024 GB |
| 8x A100 | 128 | 8 | 2,048 GB |

**Note:** The Saturn Cloud configurations shown above use A100 GPUs. H100 hours may be delivered through a different mechanism depending on the specific grant. Confirm GPU type and platform during onboarding.

- Resources **auto-shutoff after 1 hour of inactivity** to conserve your hours
- You can also create **Dask clusters** for distributed computing, connect to data sources, and link Git repositories

---

## What You Get Beyond Compute

- **No IP claims** — NVIDIA has no intention of owning IP developed under the program
- **Support letters** — NVIDIA can provide letters of support for your external grant applications (e.g., NSF, NIH, ERC)
- **GTC networking** — opportunities to present at NVIDIA's GPU Technology Conference
- **Marketing promotion** — potential visibility through NVIDIA channels
- **Unrestricted gift** — awards are structured as unrestricted gifts to the institution

---

## Post-Award Obligations

- **Acknowledge NVIDIA Corporation** in any resulting publications
- **Provide progress updates** every six months
- **Submit proposals to present** research outcomes at GTC or other NVIDIA events
- **Submit results** through the program portal (required before you can reapply)

---

## Tips for a Strong Proposal

- **Demonstrate deep knowledge of NVIDIA tools** — reviewers look for a clear understanding of how NVIDIA technology (CUDA, NeMo, Modulus, RAPIDS, etc.) will accelerate your research
- **Follow the template exactly** — non-conforming submissions are rejected without review
- **Align tightly with the CFP** — generic AI proposals that do not match the specific call will be declined
- **Be specific about resource needs** — explain why you need the requested GPU hours and how you will use them within the 6-month window
- **Check the example proposal** — NVIDIA provides a sample high-quality submission on the program portal
- **Watch the Insider Tips webinar** — it covers common mistakes and advice from actual reviewers

---

## Related NVIDIA Academic Programs

| Program | For Whom | Award |
|---|---|---|
| **[Graduate Fellowship Program](https://research.nvidia.com/graduate-fellowships)** | PhD students (completed 1st year) | Up to $60,000 + mandatory summer internship |
| **[Applied Research Accelerator Program](https://resources.nvidia.com/en-us-gps-ai-capacity-building/applied-research-pro)** | Researchers with deployment-ready projects | Technical guidance, hardware, and funding |
| **[NVIDIA Developer Program](https://developer.nvidia.com/)** | Any developer or researcher | Free access to SDKs, tools, training, and early-access programs |

---

## Additional Tips

- **This is not a self-service program** — you cannot simply sign up and get GPU hours. It is a competitive grant with a formal review process and quarterly decision cycles
- **Plan ahead** — from submission to receiving GPU access can take 3-6 months depending on the quarterly cycle
- **Students should work with their advisor** — while students cannot apply directly, they can be named as project participants
- **Join the NVIDIA Developer Program** to get notified about new CFP announcements
- **Contact:** For grant-specific questions, email AcademicGrants@nvidia.com

---

**Sources:**
- [NVIDIA Academic Grant Program](https://www.nvidia.com/en-us/industries/higher-education-research/academic-grant-program/)
- [Academic Grants Portal](https://academicgrants.nvidia.com/academicgrantprogram/s/Application)
- [Academic Grants FAQ](https://academicgrants.nvidia.com/academicgrantprogram/s/faq)
- [NVIDIA Higher Education & Research Resources](https://developer.nvidia.com/higher-education-and-research)
- [Saturn Cloud NVIDIA Grants Overview](https://saturncloud.io/nvidia-grants/overview/)
- [Insider Tips Webinar](https://www.nvidia.com/en-us/on-demand/session/other25-nvidia-agp/)
- [NVIDIA Graduate Fellowship Program](https://research.nvidia.com/graduate-fellowships)
- [University of Utah — NVIDIA Academic Grant Program](https://www.research.utah.edu/funding-item/nvidia-academic-grant-program/)

## Submission 54: Novita AI - $10 Free Credits for New Users

- Category: AI API Free Tiers
- Source Link: https://novita.ai/
- Amount: $10 in free credits
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

New users signing up for Novita AI receive **$10 in free credits** to explore 200+ AI model APIs, including LLMs, image generation, video generation, audio, and embeddings. The platform offers an **OpenAI-compatible API**, so you can switch from OpenAI with minimal code changes. No credit card is required to sign up. After credits run out, billing is strictly pay-as-you-go with no subscriptions or hidden fees.

---

## Registration (Step-by-Step)

1. Go to [novita.ai](https://novita.ai) and click **"Sign Up"** or **"Get Started"**
2. Create an account using **Google**, **GitHub**, or **email** (a new account is automatically created on first login with Google/GitHub)
3. Your **$10 in free credits** are applied to your account automatically upon signup (promotional offer; the baseline welcome voucher is $0.50, but current promotions bring this to $10)
4. Navigate to **Key Management** in account settings to create your first API key
5. Start making API calls or use the **Playground** at [novita.ai/model-api/playground](https://novita.ai/model-api/playground) to test models with no code

**Important:**
- No credit card is required to sign up or claim the free credits
- The $10 promotional credit amount may vary over time; check the current offer on the signup page
- To continue using the platform after credits are exhausted, add funds via the Billing and Payments section
- Setting up **Automatic Top-up** is recommended to avoid service interruption once you move to paid usage

---

## Available AI Models

Novita AI provides access to **200+ models** across multiple categories via a unified API.

### Large Language Models (LLMs)

| Model | Input (per M tokens) | Output (per M tokens) | Context |
|---|---|---|---|
| **DeepSeek V3.2** | $0.269 | $0.40 | — |
| **DeepSeek R1 0528** | $0.70 | $2.50 | — |
| **DeepSeek V3 Turbo** | $0.40 | $1.30 | — |
| **Qwen3 Next 80B** | $0.15 | $1.50 | — |
| **Qwen3 235B A22B** | $0.09 | $0.58 | — |
| **Qwen2.5 7B Instruct** | $0.07 | $0.07 | — |
| **Llama 3.1 8B Instruct** | $0.02 | $0.05 | — |
| **Llama 3.3 70B Instruct** | $0.135 | $0.40 | — |
| **GLM-4.7** | $0.60 | $2.20 | 204,800 |
| **GLM-4.7 Flash** | $0.07 | $0.40 | 204,800 |
| **MiniMax M2.1** | $0.30 | $1.20 | 204,800 |
| **Baidu ERNIE 4.5 21B** | $0.07 | $0.28 | — |

### Free Models (No Credits Required)

Novita AI offers several models completely free of charge:
- **Llama 3.2 1B Instruct**
- **Qwen2.5 7B Instruct**
- **GLM-4-9B-0414**
- **GLM-Z1-9B-0414**
- **BGE-M3** (multilingual embedding model)

### Image Generation

- Text-to-image starting at **$0.001/image** (standard) to **$0.005/image**
- Flux series (Kontext, 2.0 variants), Stable Diffusion, and more
- Specialized tools: background removal ($0.017/image), inpainting, upscaling, ControlNet, LoRA support

### Video Generation

- Models from Kling, Wan, Vidu, Runway, Minimax Hailuo, SeeDance
- Pricing ranges from **$0.09 to $2.34 per video** depending on model, duration, and resolution

### Audio & Speech

- Text-to-speech: **$15/M characters** (standard) to **$100/M characters** (premium MiniMax)
- Voice cloning: **$0.10/voice** (Fish Audio)
- Speech recognition: GLM ASR

### Embeddings & Reranking

- BGE-M3 (free), plus various paid embedding and reranker models

---

## OpenAI API Compatibility

Novita AI endpoints are **fully compatible with the OpenAI API standard**. This means:
- You can use the **OpenAI Python SDK** by changing only the `base_url` and `api_key`
- Existing applications built for OpenAI can switch to Novita with minimal changes
- The platform also supports **Anthropic SDK-compatible access** for integration with Claude Code-style agentic workflows
- Works with popular frameworks: **LangChain**, **LiteLLM**, **LobeChat**, **OpenCode**, and others

---

## SDKs & Developer Tools

- **Python SDK** and **JavaScript SDK** available (install via pip/npm)
- SDKs cover LLMs, image generation (Txt2Img, Img2Img), video generation, background removal, ControlNet, VAE, LoRA, and more
- **Playground** at [novita.ai/model-api/playground](https://novita.ai/model-api/playground) lets you test models in the browser with no code and auto-generates API snippets
- API key authentication via Bearer token in request headers
- Available on **Hugging Face** and **POE** for additional access points

---

## GPU Instances

Beyond the model API, Novita AI also offers GPU cloud instances (these use separate billing, not the $10 API credits):

| Feature | Details |
|---|---|
| **Available GPUs** | NVIDIA A100, H100, and others |
| **On-Demand Pricing** | ~$2.89/hour (H100) |
| **Spot Instances** | 40-60% cheaper than on-demand (interruptible) |
| **Startup Time** | Under 30 seconds |
| **Serverless GPU** | Pay-per-use, no idle costs |
| **Agent Sandbox** | ~200 ms startup for agent workloads |

---

## Referral Program

Novita AI runs a **"Give $10, Earn $10"** referral program:
- Share your referral link with a friend
- When they sign up, **both of you receive $10 in LLM API credits**
- You can earn up to **$500 total** in referral credits
- Credits apply to all Model API usage

This is an effective way to extend your free credits well beyond the initial $10.

---

## Batch Inference

For cost-sensitive workloads, Novita AI offers **batch inference** at an introductory **50% discount** on both input and output tokens for supported models. This is ideal for processing large datasets where real-time response is not needed.

---

## Additional Tips

- **No credit card required** to sign up and start using the platform, which is unusual among API providers and a significant advantage for quick experimentation
- **$10 in credits goes far** with cheaper models: at $0.02/M input tokens for Llama 3.1 8B, $10 buys roughly 500M input tokens
- **Free models exist** (Llama 3.2 1B, Qwen2.5 7B, GLM-4-9B, BGE-M3) that cost nothing at all, even after your credits run out
- **The Playground is free** for browser-based testing without needing to write code
- **Pricing is competitive** compared to direct providers; DeepSeek V3.2 at $0.269/M input is in line with or cheaper than other inference providers
- **Support is responsive** via Discord; users report that Novita will add custom models on request
- **Global availability** with no region restrictions reported
- **Trustpilot reviews** are generally positive, praising fast API integration, low pricing, and model variety

---

**Sources:**
- [Novita AI Homepage](https://novita.ai)
- [Novita AI Pricing](https://novita.ai/pricing)
- [Novita AI Model Library](https://novita.ai/models)
- [Novita AI Quickstart Guide](https://novita.ai/docs/guides/quickstart)
- [Novita AI Referral Program](https://novita.ai/referral)
- [Novita AI Playground](https://novita.ai/model-api/playground)
- [Novita AI Free Models Blog Post](https://blogs.novita.ai/free-model/)
- [Novita AI DeepSeek R1 0528 Free Credits](https://blogs.novita.ai/deepseek-r1-0528-now-available-on-novita-ai-claim-your-10-free-credits/)
- [Novita AI on GitHub](https://github.com/novitalabs)
- [Novita AI GPU Rental Blog](https://blogs.novita.ai/rent-cheap-a100-and-h100-boost-training-efficiency-with-novita-ai/)
- [Novita AI Trustpilot Reviews](https://www.trustpilot.com/review/novita.ai)

## Submission 53: Modal Startup Credits - Up to $25,000

- Category: Startup Programs
- Source Link: https://modal.com/startups
- Amount: Up to $25,000 (varies by tier: $5K bootstrapped, $25K seed-to-Series-A, custom for Series B+)
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Modal offers **up to $25,000 in free serverless GPU compute credits** for early-stage AI startups through its startup program. Credits cover GPU, CPU, and memory usage on Modal's serverless platform, which handles AI inference, fine-tuning, batch processing, and training workloads. Beyond credits, accepted startups get direct access to Modal's engineering team, go-to-market amplification, and an exclusive founder network. Modal charges per second with no idle costs and scales from zero to thousands of GPUs automatically.

---

## Registration (Step-by-Step)

1. Go to [modal.com/signup](https://modal.com/signup)
2. Sign up using **GitHub**, **Google**, or **SSO** — no credit card required at signup
3. You immediately receive **$30/month in free compute credits** on the Starter plan
4. To apply for startup credits, go to [modal.com/startups](https://modal.com/startups)
5. Click **"Apply Now"** and fill out the application form with your company details, funding stage, and use case
6. Wait for review — Modal evaluates applications based on your tier eligibility and VC backing
7. If approved, credits are applied to your workspace as a one-time grant

**Important:**
- No credit card is required to sign up and start using Modal with the $30/month free tier
- A payment method is needed once you exceed free credits or want to scale beyond the free tier
- Startup credits are a **one-time grant** — you cannot reapply after receiving credits in a given tier
- You can also transact through the **AWS and GCP Marketplace** to use committed cloud spend on Modal

---

## Credit Tiers

Modal's startup program has three tiers based on your funding stage:

| Tier | Credit Amount | Requirements |
|---|---|---|
| **Bootstrapped** | Up to **$5,000** | Full-time team, less than $500K raised outside Modal's VC network, active company website |
| **Seed to Series A** | Up to **$25,000** | VC funding up to $30M, investor from Modal's partner network |
| **Series B+ / Scaling** | **Custom** | Over $30M in funding or post-Series B, partner VC investor |

**Key detail:** The Seed-to-Series-A and Series B+ tiers require an investor from Modal's **partner VC network**. If your investors are not in that network, you may only qualify for the Bootstrapped tier. Modal's known investors include **Amplify Partners**, **Redpoint**, and **Left Lane Capital**, but the full partner network is not publicly listed.

---

## What Credits Cover

Credits apply to all Modal compute usage, billed per second with no minimums:

### GPU Pricing (per second)

| GPU | Per Second | ~Hourly Equivalent |
|---|---|---|
| **NVIDIA B200** | $0.001736 | ~$6.25/hr |
| **NVIDIA H200** | $0.001261 | ~$4.54/hr |
| **NVIDIA H100** | $0.001097 | ~$3.95/hr |
| **NVIDIA A100 (80 GB)** | $0.000694 | ~$2.50/hr |
| **NVIDIA A100 (40 GB)** | $0.000583 | ~$2.10/hr |
| **NVIDIA L40S** | $0.000542 | ~$1.95/hr |
| **NVIDIA A10G** | $0.000306 | ~$1.10/hr |
| **NVIDIA L4** | $0.000222 | ~$0.80/hr |
| **NVIDIA T4** | $0.000164 | ~$0.59/hr |

### CPU & Memory

| Resource | Per Second |
|---|---|
| **CPU** (physical core) | $0.0000131 |
| **Memory** (per GiB) | $0.00000222 |

With $25,000 in credits, you could run roughly **6,300 hours of H100 GPU time**, **11,900 hours of A100 (40 GB)**, or **42,500 hours of T4 GPU time**.

---

## Platform Features

Modal is a **serverless compute platform** — there are no instances to manage, no Kubernetes, no Docker configuration. Key features:

- **Sub-second cold starts** — containers spin up in under a second using Modal's custom Rust-based container runtime
- **Auto-scaling** — scales from zero to thousands of GPUs based on traffic, scales back to zero when idle (no idle costs)
- **Code-first infrastructure** — GPU type, container environment, and dependencies are all defined in Python code (no YAML, no config files)
- **Web endpoints** — deploy inference APIs with streaming, authentication, and configurable timeouts
- **Volumes and storage** — persistent volumes, cloud bucket mounts, dicts, and queues for data sharing
- **Secrets management** — built-in secrets and environment variable handling
- **Multi-language SDKs** — Python (primary), TypeScript, and Go
- **Sandboxes and notebooks** — run AI-generated code in isolated environments, GPU-backed notebooks

---

## Beyond Credits: Startup Program Benefits

The startup program includes more than just compute credits:

- **Engineering support** — direct access to Modal's engineering team for architecture guidance, debugging, and optimization
- **GTM amplification** — Modal promotes your launches and fundraising through their channels
- **Founder network** — exclusive community of AI startup founders with online and in-person events
- **Technical onboarding** — help getting your workloads running efficiently on Modal's platform

---

## Free Tier (No Application Needed)

Even without the startup program, every Modal account gets:

| Feature | Starter Plan (Free) |
|---|---|
| **Monthly credits** | $30/month |
| **Workspace seats** | 3 |
| **Max containers** | 100 |
| **GPU concurrency** | 10 |
| **Cost** | $0/month |

The $30/month free tier is enough to experiment with Modal, run small inference jobs, and prototype before applying for startup credits.

---

## Additional Tips

- **No credit card needed to start** — sign up and use the $30/month free tier without entering payment details. A payment method is only required once you exceed free credits
- **Per-second billing with no minimums** — you only pay for actual compute time. If your GPU runs for 3 seconds, you pay for 3 seconds. No hourly minimums, no reservations
- **Serverless sweet spot** — Modal is most cost-effective when your GPU utilization is bursty or below 40%. If you need 24/7 dedicated GPUs, reserved instances elsewhere may be cheaper
- **AWS/GCP Marketplace** — you can pay for Modal through AWS or GCP committed spend, which may count toward existing cloud credits from other startup programs
- **Modal for Academics** — a separate program offering up to $10,000 in credits for academic research
- **Team plan** — $250/month with $100/month in included credits, unlimited seats, 1,000 containers, and 50 GPU concurrency. Worth considering once you outgrow the Starter plan
- **Modal raised $87M Series B in September 2025** at a $1.1B valuation, indicating strong platform stability and continued investment

---

**Sources:**
- [Modal Startups Program](https://modal.com/startups)
- [Modal Pricing](https://modal.com/pricing)
- [Modal Documentation Guide](https://modal.com/docs/guide)
- [Modal Signup](https://modal.com/signup)
- [Modal Billing Docs](https://modal.com/docs/guide/billing)
- [Modal for Academics](https://modal.com/academics)
- [Contrary Research — Modal Business Breakdown](https://research.contrary.com/company/modal)

## Submission 52: Modal - $30/Month Free Compute Credits

- Category: AI API Free Tiers
- Source Link: https://modal.com/pricing
- Amount: $30/month in free compute credits
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Modal gives every Starter plan user **$30 in free compute credits each month** with no monthly fee and no credit card required. You pay only for actual compute time — by the CPU cycle, billed per second — so bursty workloads like inference, batch jobs, and fine-tuning cost a fraction of what reserved instances charge. Modal supports a wide range of NVIDIA GPUs (T4 through B200), serverless functions, scheduled cron jobs, web endpoints, and persistent volumes. Credits reset monthly and never roll over.

---

## Registration (Step-by-Step)

1. Go to [modal.com/signup](https://modal.com/signup)
2. Click **"Continue with GitHub"**, **"Continue with Google"**, or use **SSO** — no email/password form, just OAuth
3. Authorize the Modal app (GitHub requests `user:email` and optionally `read:org` for workspace invites — it does **not** access your repos)
4. Your workspace is created immediately with $30 in free monthly credits
5. Install the Python package locally: `pip install modal`
6. Run `modal setup` in your terminal — this opens a browser tab to authenticate your CLI with a token
7. You are ready to deploy — run `modal run your_script.py` to test

**Important:**
- **No credit card required** for the Starter plan. You only need to add billing if you exceed $30/month
- The $30 credit resets every month — unused credits do **not** roll over
- One workspace is created per signup; you can invite up to **3 seats** on the Starter plan
- Modal runs on top of major cloud providers — you do not need your own AWS/GCP account

---

## Compute Pricing (What $30 Gets You)

Modal charges per second with no minimum commitment. Here is what the main resources cost:

### GPU Pricing

| GPU | Per Second | Approx. Per Hour | ~Free Hours/Month ($30) |
|---|---|---|---|
| **NVIDIA T4** (16 GB) | $0.000164 | $0.59 | ~50 h |
| **NVIDIA L4** (24 GB) | $0.000222 | $0.80 | ~37 h |
| **NVIDIA A10G** (24 GB) | $0.000306 | $1.10 | ~27 h |
| **NVIDIA L40S** (48 GB) | $0.000542 | $1.95 | ~15 h |
| **NVIDIA A100 40 GB** | $0.000583 | $2.10 | ~14 h |
| **NVIDIA A100 80 GB** | $0.000694 | $2.50 | ~12 h |
| **NVIDIA H100** (80 GB) | $0.001097 | $3.95 | ~7.5 h |
| **NVIDIA H200** (141 GB) | $0.001261 | $4.54 | ~6.6 h |
| **NVIDIA B200** (192 GB) | $0.001736 | $6.25 | ~4.8 h |

### CPU & Memory

| Resource | Per Second | Approx. Per Hour |
|---|---|---|
| **CPU** | $0.0000131 / core | $0.047 / core |
| **Memory** | $0.00000222 / GiB | $0.008 / GiB |
| **Sandbox/Notebook CPU** | $0.00003942 / core | $0.142 / core |
| **Sandbox/Notebook Memory** | $0.00000672 / GiB | $0.024 / GiB |

Minimum CPU allocation is 0.125 cores. There are **no idle charges** — when your function is not running, you pay nothing.

---

## Starter Plan Limits

The free Starter plan has the following concurrency and deployment limits:

| Resource | Starter Limit |
|---|---|
| **Concurrent containers** | 100 |
| **GPU concurrency** | 10 |
| **Deployed cron jobs** | 5 |
| **Deployed web endpoints** | 8 |
| **Log retention** | 1 day |
| **Workspace seats** | 3 |
| **Deployment rollback versions** | 3 |
| **Custom domains** | Not available |

These limits are sufficient for individual developers and small teams running experiments, inference endpoints, or periodic batch jobs.

---

## Key Platform Features

- **Serverless Functions** — decorate any Python function with `@app.function()` and Modal handles containerization, scaling, and GPU attachment automatically
- **Per-Second Autoscaling** — scales from zero to hundreds of GPUs in seconds, then back to zero when idle
- **Sub-Second Cold Starts** — custom Rust-based runtime starts GPU containers in ~1-2 seconds
- **Web Endpoints** — expose functions as HTTPS endpoints with `@app.web_endpoint()` for serving models or APIs
- **Cron Jobs** — schedule recurring tasks with `@app.cron()` syntax
- **Volumes** — persistent distributed storage (up to 1 TiB per file, ~2.5 GB/s bandwidth) for model weights, datasets, and artifacts
- **Cloud Bucket Mounts** — mount S3 or GCS buckets directly as file systems (Cloudflare R2 recommended for zero egress fees)
- **Secrets** — securely inject API keys and credentials as environment variables
- **Sandboxes** — run untrusted or AI-generated code in isolated containers
- **Notebooks** — GPU-backed Jupyter notebooks with zero cold-boot penalty
- **Infrastructure as Code** — define container images, GPU requirements, and dependencies in Python — no YAML, no Dockerfiles required

---

## Plan Comparison

| Feature | Starter ($0/mo) | Team ($250/mo) | Enterprise (Custom) |
|---|---|---|---|
| Free monthly credits | $30 | $100 | Custom |
| Containers | 100 | 1,000 | Custom |
| GPU concurrency | 10 | 50 | Custom |
| Cron jobs | 5 | Unlimited | Unlimited |
| Web endpoints | 8 | Unlimited | Unlimited |
| Log retention | 1 day | 30 days | Custom |
| Seats | 3 | Unlimited | Unlimited |
| Custom domains | No | Yes | Yes |
| Static IP proxy | No | Yes | Yes |
| Okta SSO | No | No | Yes |
| Support | Community | Community | Private Slack + ML engineering |

---

## Startup & Academic Programs

### Startups (up to $25,000 in credits)

| Tier | Credits | Requirements |
|---|---|---|
| **Bootstrapped** | Up to $5,000 | Full-time team, raised < $500K outside Modal's VC network, active website |
| **Seed to Series A** | Up to $25,000 | Raised up to $30M via Modal's partner VC network (YC, a16z, Sequoia, Khosla, etc.) |
| **Series B+** | Custom | Raised > $30M or post-Series B, VC from partner network |

Apply at [modal.com/startups](https://modal.com/startups). Credits are granted once per tier.

### Academics (up to $10,000 in credits)

Graduate students, labs, and researchers can apply for up to **$10,000 in free compute credits** for research projects. Credits typically expire after conference submission results are finalized and are granted once per conference. Educators should contact partnerships@modal.com for dedicated teaching programs.

Apply at [modal.com/academics](https://modal.com/academics).

---

## Additional Tips

- **No billing surprise risk on Starter** — if you exceed $30 without a payment method, your workloads stop rather than incurring charges. You must explicitly add billing to go beyond the free tier
- **AWS/GCP Marketplace** — if you have committed cloud spend, you can route Modal usage through the AWS or GCP Marketplace to draw from existing commitments
- **Model weights workflow** — store large model files in a Modal Volume and mount them into functions to avoid re-downloading on each cold start
- **Cloudflare R2 for datasets** — Modal recommends R2 over S3 for large dataset storage because Cloudflare charges zero egress fees
- **Python-first, but not Python-only** — the primary SDK is Python, but JavaScript/TypeScript and Go SDKs are available for calling functions, managing sandboxes, and interacting with resources
- **No Docker required locally** — Modal builds container images in the cloud from your Python code; you never need to write a Dockerfile
- **Compare with alternatives** — for always-on workloads running 24/7, reserved GPU instances (RunPod, Lambda, etc.) may be cheaper; Modal shines for bursty, intermittent, or autoscaling workloads

---

**Sources:**
- [Modal Pricing](https://modal.com/pricing)
- [Modal Signup](https://modal.com/signup)
- [Modal Documentation Guide](https://modal.com/docs/guide)
- [Modal Account Setup](https://modal.com/docs/guide/modal-user-account-setup)
- [Modal for Startups](https://modal.com/startups)
- [Modal for Academics](https://modal.com/academics)
- [Modal Volumes Documentation](https://modal.com/docs/guide/volumes)
- [Modal on AWS Marketplace](https://aws.amazon.com/marketplace/pp/prodview-j727623xqhh2k)

## Submission 51: Mistral AI La Plateforme - Free Tier

- Category: AI API Free Tiers
- Source Link: https://console.mistral.ai/
- Amount: Rate-limited free usage (~1 RPS, ~500K tokens/min, up to ~1B tokens/month per model)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Mistral AI offers a **free Experiment plan** on La Plateforme (their serverless API platform) that lets you use all Mistral models for experimentation and prototyping. **No credit card required** — just a verified phone number. Rate limits are restrictive (approximately **1 request/second per model**, **500K tokens/minute**, up to **1B tokens/month**), but generous enough for prototyping and evaluation. The main catch: **your API inputs and outputs may be used to train Mistral's models** unless you manually opt out in the Admin Console.

---

## Registration (Step-by-Step)

1. Go to [console.mistral.ai](https://console.mistral.ai)
2. Click **"Sign up"** — you can register with **email/password**, **Google**, or **GitHub**
3. **Verify your email** address via the confirmation link sent to your inbox
4. Once logged in, go to the **Admin Console** (accessible from the main menu in Mistral AI Studio)
5. Navigate to **Subscription** and select the **Mistral AI Studio** tab
6. Click **"Compare plans"**, then click **"Experiment for free"**
7. Review the plan details, accept the **Terms of Service** and **Privacy Policy**, then click **Subscribe**
8. **Verify your phone number** — enter a valid phone number and confirm via SMS code
9. Your Experiment plan is now active. Navigate to **API Keys** at [console.mistral.ai/home](https://console.mistral.ai/home) and click **"Create new secret key"** to generate your API key

**Important:**
- One phone number per Experiment plan — you cannot reuse the same phone number for multiple accounts
- No credit card is required at any point
- You can start sending API requests immediately after setup

---

## Available Models

The Experiment plan provides access to Mistral's full model lineup. All models listed below are available via the API at their standard per-token pricing (which on the Experiment plan costs you nothing — you just hit rate limits instead of being billed).

### Premier / Frontier Models

| Model | Parameters | Context Window | Notes |
|---|---|---|---|
| **Mistral Large 3** (2512) | 41B active / 675B total (MoE) | 262K tokens | Most capable model, Apache 2.0 |
| **Mistral Medium 3.1** | — | 131K tokens | GPT-4 class performance |
| **Mistral Medium 3** | — | 131K tokens | Strong general-purpose model |

### Small & Efficient Models

| Model | Parameters | Context Window | Notes |
|---|---|---|---|
| **Mistral Small 3.2** | 24B | 131K tokens | Multimodal (vision) |
| **Mistral Small 3.1** | 24B | 131K tokens | Best value small model |
| **Mistral Nemo** | 12B | 131K tokens | Cheapest model on platform |
| **Ministral 3B / 8B / 14B** | 3B–14B | 131K–262K tokens | Edge and mobile-friendly |

### Code Models

| Model | Context Window | Notes |
|---|---|---|
| **Codestral** (2508) | 256K tokens | Code completion |
| **Devstral 2** (2512) | 262K tokens | Frontier code agent model |
| **Devstral Small 2** | 24B | Optimized for consumer GPUs |

### Specialized Models

| Model | Use Case |
|---|---|
| **Magistral Small / Medium** | Reasoning (chain-of-thought) |
| **Pixtral 12B / Pixtral Large** | Vision and multimodal |
| **Voxtral Small** | Audio transcription |
| **Mistral Small Creative** | Creative writing |
| **Saba** | Middle-Eastern and South-Asian languages |

---

## Rate Limits (Experiment Plan)

Rate limits on the Experiment plan are **restrictive by design** — they are intended for testing and prototyping, not production. Limits are set **per model** at the workspace level.

| Metric | Approximate Limit |
|---|---|
| **Requests per second (RPS)** | ~1 per model |
| **Tokens per minute** | ~500,000 per model |
| **Tokens per month** | Up to ~1B (varies by model) |

**Notes:**
- Token limits include both input and output tokens combined
- Exact limits vary by model and may change — check your current limits at [admin.mistral.ai/plateforme/limits](https://admin.mistral.ai/plateforme/limits)
- The same inference engine runs for free and paid tiers; only concurrency and caps differ
- If you hit rate limits, you receive HTTP 429 errors — requests are not queued

---

## Data Training & Privacy

This is the biggest caveat of the free tier:

- **By default, your API inputs and outputs are used to train Mistral's models** on the Experiment plan
- You **can opt out**, but it requires manual action

### How to Opt Out of Data Training

1. Log in to the **Admin Console** at [admin.mistral.ai](https://admin.mistral.ai)
2. Click **Privacy** in the left-hand navigation bar
3. Find the **"Anonymous improvement data"** section
4. **Disable the toggle** to prevent your API calls from being used for training

**Comparison by plan:**

| Plan | Data Used for Training? |
|---|---|
| **Experiment (Free)** | Yes, by default (opt-out available) |
| **Scale (Paid)** | No, opted out by default |
| **Team / Enterprise** | No, opted out by default |

---

## Features Included in Free Tier

| Feature | Available? |
|---|---|
| **Chat completions API** | Yes |
| **Embeddings API** | Yes |
| **Function calling** | Yes |
| **JSON mode** | Yes |
| **Vision (multimodal)** | Yes (with supported models) |
| **Code completion (FIM)** | Yes (Codestral) |
| **Flash Answers** | Yes |
| **Code interpreter** | Yes |
| **Document uploads** | Yes |
| **Agents API** | Limited (full access requires Scale plan) |
| **Fine-tuning** | No (requires Scale plan) |
| **Batch API** | No (requires Scale plan) |

---

## Upgrading to Scale Plan

When you outgrow the Experiment plan, the **Scale plan** is Mistral's pay-as-you-go tier:

- **No monthly subscription fee** — pure token-based billing
- **Credit card required** to activate
- **Higher rate limits** (specific limits shown in your admin dashboard)
- **Data not used for training** by default
- **Full access** to agents, fine-tuning, batch API
- Upgrade is seamless — your API keys and endpoints stay the same

Example pricing on Scale plan (per 1M tokens):

| Model | Input | Output |
|---|---|---|
| Mistral Nemo | $0.02 | $0.04 |
| Mistral Small 3.1 | $0.03 | $0.11 |
| Mistral Small 3.2 | $0.06 | $0.18 |
| Codestral | $0.30 | $0.90 |
| Mistral Medium 3 | $0.40 | $2.00 |
| Mistral Large 3 | $0.50 | $1.50 |

---

## Additional Tips

- **Le Chat vs. La Plateforme** — Le Chat (chat.mistral.ai) is Mistral's consumer chatbot with its own free tier (~25 messages/day). La Plateforme (console.mistral.ai) is the developer API platform. They are separate products with separate accounts and limits
- **OpenAI-compatible API** — Mistral's API follows the OpenAI API format, so most OpenAI client libraries work with minimal changes (just swap the base URL and API key)
- **European data residency** — Mistral AI is a French company. Data processing occurs in the EU, which may be relevant for GDPR compliance
- **Open-weight models** — Many Mistral models (including Mistral Large 3, Small 3.2, Nemo) are released under Apache 2.0 license. You can self-host them for free if you have the hardware, bypassing all API rate limits
- **Context windows are generous** — even on the free tier, models like Mistral Large 3 support 262K token context windows
- **No spending caps concern** — since the Experiment plan is truly free with rate limits (not credits), you cannot accidentally get billed. Billing only starts if you explicitly upgrade to Scale and add payment info

---

**Sources:**
- [Mistral AI Console](https://console.mistral.ai)
- [Mistral AI Pricing](https://mistral.ai/pricing)
- [Mistral AI Rate Limits & Usage Tiers](https://docs.mistral.ai/deployment/ai-studio/tier)
- [How to Try the API Free with Experiment Plan](https://help.mistral.ai/en/articles/455206-how-can-i-try-the-api-for-free-with-the-experiment-plan)
- [Opt Out of Data Training](https://help.mistral.ai/en/articles/455207-can-i-opt-out-of-my-input-or-output-data-being-used-for-training)
- [Mistral AI Models Documentation](https://docs.mistral.ai/getting-started/models)
- [Mistral AI API Rate Limits FAQ](https://help.mistral.ai/en/articles/424390-how-do-api-rate-limits-work-and-how-do-i-increase-them)
- [Mistral 3 Announcement](https://mistral.ai/news/mistral-3)
- [Mistral AI API Pricing (pricepertoken.com)](https://pricepertoken.com/pricing-page/provider/mistral-ai)

## Submission 50: Microsoft for Startups - YC Partnership ($350,000)

- Category: Startup Programs
- Source Link: https://www.microsoft.com/en-us/startups
- Amount: $350,000 in Azure credits + $2,500 OpenAI credits + software licenses
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Y Combinator batch companies get **$350,000 in Azure credits** through a special Microsoft partnership launched in November 2023. This is one of the largest cloud credit packages available to YC startups -- significantly more than the standard $150,000 (now $100,000) investor-track offer. Credits cover Azure compute, data services, analytics, AI workloads, and priority access to GPU infrastructure for AI model training. As of mid-2024, **58% of YC startups** had accepted the offer, up from just 5% building on Azure before the partnership.

---

## How to Access (YC Companies)

1. **Get accepted into a Y Combinator batch** -- this offer is exclusive to active YC batch companies
2. YC provides a **referral code** for the Microsoft for Startups Investor Network
3. Go to [startups.microsoft.com](https://www.microsoft.com/en-us/startups) and click **"Get started now"**
4. Enter the **YC referral code** when prompted
5. Complete the application form -- you will need a **LinkedIn account** and basic company details (name, stage, description)
6. Select your development stage (Concept design, Prototyping, Building MVP, MVP in market, or Established market)
7. Wait for approval -- average review time is **3 business days**, notification comes via email
8. Once approved, log in to the [Microsoft for Startups Portal](https://portal.startups.microsoft.com) and click **"Activate Azure"**
9. Link your Microsoft account and complete Azure sign-up (requires a payment method on file, but you will not be charged while credits are active)
10. Credits may take up to **24 hours** to appear on your account

**Important:**
- The referral code comes through YC -- you cannot get it from Microsoft directly
- You must be in an **active YC batch** (not just a YC alum applying independently)
- If declined, you can reapply after **14 days**
- Applications cannot be edited after submission; you can delete and reapply if still pending

---

## What the Credits Cover

The $350,000 in Azure credits can be applied toward **any Azure product** listed on [azure.microsoft.com/products](https://azure.microsoft.com/products/), including:

| Category | Examples |
|---|---|
| **Compute** | Virtual Machines, App Service, Container Apps, Kubernetes (AKS) |
| **AI / ML** | Azure OpenAI Service, Azure AI Foundry, Azure Machine Learning, Cognitive Services |
| **GPU Infrastructure** | ND A100 VMs for model training and fine-tuning (priority access for YC companies) |
| **Data & Analytics** | Cosmos DB, PostgreSQL, SQL Database, Synapse Analytics, Data Factory |
| **Storage & Networking** | Blob Storage, CDN, Virtual Network, Load Balancer |
| **Developer Tools** | Azure DevOps, Azure Functions, Logic Apps |

**Credits cannot be used for:**
- Microsoft Azure Marketplace (third-party products)
- Microsoft Support Plans
- Products sold separately from Azure
- Credits cannot be combined with other licensing, prepaid, or credit offers

---

## GPU Access for AI Workloads

A major differentiator of the YC partnership is **priority access to GPU infrastructure**:

- YC startups get access to **ND A100 Azure VM clusters** configured for model training and fine-tuning
- This includes optimization technologies: **LoRA (Low Rank Adaptation)**, **DeepSpeed**, and **ONNX Runtime**
- Access is **time-bound** -- intended for testing and trialing models, not indefinite production use
- Startups also get **1:1 guidance from Microsoft AI experts** on architecture, performance, and implementation
- Azure OpenAI Service access (GPT-5 mini/nano, GPT-4.1, GPT-4o, o4-mini, o3-mini, etc.) is available and credits apply to API usage. Advanced models (GPT-5.2, GPT-5.1, GPT-5, o3) require [limited-access registration](https://aka.ms/oai/gpt5access)

---

## Credit Duration and Expiration

- Credits are valid for **up to 2 years** from activation
- Once credits are exhausted or expire, your Azure account transitions to **pay-as-you-go** pricing
- A payment method is required on file during activation, but you are **not charged** while credits remain
- Unused credits **do not roll over** or extend beyond the expiration date
- The lifetime cap for free Azure credits through this program is **$350,000** -- once reached, no additional credits are available

---

## Additional Benefits (Beyond Azure Credits)

The YC partnership includes the full Microsoft for Startups benefit package:

| Benefit | Details |
|---|---|
| **OpenAI Credits** | $2,500 in OpenAI API credits for experimentation |
| **GitHub Enterprise** | 20 user seats |
| **Microsoft 365 Business Premium** | 50 seats |
| **Visual Studio Enterprise** | 5 user licenses |
| **LinkedIn Premium** | Included |
| **Microsoft Clarity** | Free analytics tool |
| **Dynamics 365** | 10-25 seats depending on module |
| **Power Platform** | Power Apps, Power Automate, Power BI -- 10 users each |
| **Azure Standard Support** | 24/7 expert assistance at no cost |
| **Go-to-Market Support** | Value proposition refinement, Microsoft channel sales access |

---

## Eligibility Requirements

To qualify for the $350,000 YC package, your startup must:

- Be an **active Y Combinator batch company** (the referral code comes through YC)
- Have a **software-based product** as core business (owned, not licensed)
- Have received fewer than **$350,000 in lifetime free Azure credits**
- Be headquartered in a country where **Azure services are available**
- Have **not raised Series C funding** or later
- Be **privately held** and **for-profit**
- **Not be** an educational institution, government org, consultancy, agency, dev shop, or crypto mining operation

---

## 2025 Program Changes -- What to Know

As of July 2025, Microsoft overhauled the broader Startups program:

- The old "Founders Hub" brand is retired; everything is now **Microsoft for Startups**
- The general investor-track offer starts at **$100,000** (down from $150,000), with the opportunity to earn more based on engagement
- Non-investor-backed startups get a **self-service track**: $1,000 immediately, up to $5,000 after business verification
- **YC partnership credits are separate** from the general program tiers and remain at the $350,000 level for eligible batch companies
- Existing activated credits remain valid until their expiration date

---

## Competitive Context

YC companies can stack cloud credits from multiple providers:

| Provider | YC Credit Offer |
|---|---|
| **Microsoft Azure** | $350,000 in Azure credits + GPU access |
| **AWS** | $350,000 in AWS credits + $300,000 for Trainium/Inferentia AI chips |
| **Google Cloud** | Up to $350,000 for AI-focused startups |

Accepting credits from multiple providers is common and considered strategic for cost management. As one founder noted, accepting credits is essentially like raising additional capital.

---

## Additional Tips

- **Stack credits** -- most YC companies accept offers from Azure, AWS, and GCP simultaneously. There is no exclusivity requirement
- **GPU access is time-bound** -- plan your model training sprints around the GPU availability window rather than expecting indefinite access
- **Set budget alerts** in the Azure portal immediately after activation to avoid surprise charges when credits run low
- **Stopping a VM is not enough** -- you must **deallocate** it to stop incurring charges. Stopped-but-not-deallocated VMs still cost money
- **Not all models are in all regions** -- check Azure regional availability before creating resources, especially for Azure OpenAI models
- **The $2,500 OpenAI credits** are separate from the $350,000 Azure credits and are specifically for the OpenAI API (platform.openai.com), not Azure OpenAI Service
- If you are a YC alum but not in an active batch, you may still qualify for the standard investor-track offer ($100,000+) through the YC Alumni network

---

**Sources:**
- [Microsoft for Startups](https://www.microsoft.com/en-us/startups)
- [Microsoft for Startups Investor Offer Benefits](https://learn.microsoft.com/en-us/microsoft-for-startups/benefits)
- [Microsoft for Startups Application Process](https://learn.microsoft.com/en-us/microsoft-for-startups/application)
- [Azure Credit Offer Details](https://learn.microsoft.com/en-us/microsoft-for-startups/benefits/azure)
- [Microsoft Blog: Startups to Access High-Performance Azure Infrastructure](https://blogs.microsoft.com/blog/2023/11/07/startups-to-access-high-performance-azure-infrastructure-accelerating-ai-breakthroughs/)
- [CNBC: Microsoft Touts Cloud Momentum from Y Combinator Startups](https://www.cnbc.com/2024/08/02/microsoft-touts-cloud-momentum-from-y-combinator-startups.html)
- [Changes to Microsoft for Startups Program (July 2025)](https://learn.microsoft.com/en-us/microsoft-for-startups/changes-microsoft-for-startups)
- [Evolve ETFs: How Microsoft is Winning AI Startups](https://evolveetfs.com/2024/09/how-microsoft-is-winning-ai-startups-with-350k-in-azure-cloud-credits/)

## Submission 49: Microsoft for Startups - Investor Offer ($150,000)

- Category: Startup Programs
- Source Link: https://learn.microsoft.com/en-us/microsoft-for-startups/benefits
- Amount: Up to $150,000 in Azure credits (starting at $100,000, unlockable with qualifying usage) + Microsoft 365, Dynamics 365, Power Platform, GitHub Enterprise, and Visual Studio licenses
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Startups backed by an investor in the **Microsoft for Startups Investor Network** can receive up to **$150,000 in Azure credits**, starting at $100,000 with the ability to unlock more by demonstrating active usage of qualifying Azure services. The program also includes **50 seats of Microsoft 365 Business Premium**, **Dynamics 365** and **Power Platform** licenses, **GitHub Enterprise** (20 users), **Visual Studio Enterprise** (5 users), access to **on-demand GPU VMs**, dedicated Azure technical experts, and go-to-market support through Microsoft's enterprise sales channels. OpenAI API credits (historically $2,500) have been largely discontinued as of mid-2025 -- startups should use **Azure OpenAI** services with their Azure credits instead.

---

## Registration (Step-by-Step)

1. **Get a referral code** from your investor (VC, accelerator, incubator, or university) that is part of the Microsoft for Startups Investor Network. The list of participating investors is not public -- ask your investor directly
2. Go to [startups.microsoft.com](https://portal.startups.microsoft.com/) and click **"Get started now"**
3. Enter the **referral code** provided by your investor
4. Sign in with a **LinkedIn account** (mandatory for all applicants)
5. Fill in your **business details**: company name, website, development stage (Concept Design, Prototyping, Building MVP, MVP in Market, or Established Market)
6. Confirm your startup meets all eligibility criteria
7. Submit the application and wait for review -- typically **3 business days**
8. If approved, you receive an email notification. Log into the portal and click **"Activate Azure"** to link your Microsoft account and receive credits
9. Complete Azure sign-up (a **payment method is required**, but you will not be charged while credits are active)
10. Credits may take up to **24 hours** to appear in your Azure account

**Important:**
- Submitted applications **cannot be edited** -- you must delete and resubmit if changes are needed
- If declined, you can **reapply after 14 days**. The decline email includes reasons
- If you have referral codes from **multiple investors**, contact support with all codes
- Referral codes are **time-bound** and have expiration dates

---

## Eligibility Requirements

Your startup must meet **all** of the following criteria:

| Requirement | Details |
|---|---|
| **Investor affiliation** | Must be backed by an investor in the Microsoft for Startups Investor Network |
| **Software-based product** | Core business must be a software product or service that you own (not licensed from another party) |
| **Lifetime Azure credits** | Must have received fewer than $350,000 in lifetime free Azure credits |
| **Funding stage** | Must not have raised Series C funding or beyond |
| **Company status** | Privately held, for-profit business |
| **Headquarters** | Located in a country where Azure services are available |
| **Excluded types** | Educational institutions, government orgs, personal blogs, dev shops, consultancies, agencies, and crypto/bitcoin mining operations are not eligible |

---

## Credit Tiers and Unlocking More Credits

The investor offer does **not** give you all $150,000 at once. Credits are released in stages:

- **Starting allocation**: Most startups begin with **$100,000** in Azure credits
- **Earning more**: You can increase your allocation by spending **$100+ USD in a single month** on qualifying Azure services:
  - Azure Monitor
  - Log Analytics
  - Application Insights
  - Microsoft Defender for Cloud
  - Microsoft Sentinel
  - Microsoft Purview
- **Credit increase timeline**: Once qualifying usage criteria are met, additional credits appear within **up to 30 days**
- **Maximum cap**: Your total eligibility is capped at your offer maximum (up to $150,000) minus any credits already used
- **No rollover**: Once you reach the maximum, no additional credits are issued

**Example:** If your offer cap is $150,000 and you have already used $50,000, you can unlock up to $100,000 more.

---

## Included Licenses and Tools

| Product | Included Amount |
|---|---|
| **Microsoft 365 Business Premium** | 50 seats |
| **Dynamics 365 Customer Service Enterprise** | 10 seats |
| **Dynamics 365 Sales Enterprise** | 10 seats |
| **Dynamics 365 Team Members** | 25 seats |
| **Power Apps** | 10 users |
| **Power Automate** | 10 users |
| **Power BI** | 10 users |
| **GitHub Enterprise** | 20 users |
| **Visual Studio Enterprise** | 5 users |
| **LinkedIn Premium** | Free access |
| **Microsoft Clarity** | Free access |

---

## GPU Access and AI Workloads

The investor offer includes access to **on-demand GPU VMs** for AI workloads such as model training, fine-tuning, and inference deployment. GPU VM costs are deducted from your Azure credit balance. Available GPU VM families include:

- **NCas T4 v3** -- entry-level inference and light training
- **NC A100 v4** -- heavy training and fine-tuning
- **ND A100 v4** -- large-scale distributed training

GPU VM availability varies by Azure region. Some GPU SKUs may require a **quota increase request** through the Azure portal, which typically takes 1-3 business days.

---

## Technical and Go-to-Market Support

- **Azure Technical Experts**: Direct access to Microsoft engineers for architecture review, performance optimization, and AI implementation guidance
- **Azure Standard Support**: 24/7 expert assistance at no cost
- **Go-to-Market Support**: Help with value proposition refinement, marketing strategy, and sales through Microsoft channels
- **Microsoft Enterprise Network**: Connections to Microsoft's enterprise customer base for partnership and sales opportunities

---

## OpenAI API Credits (Largely Discontinued)

Historically, the program included **$2,500 in OpenAI API credits** (on platform.openai.com). As of mid-2025, this benefit has been **largely discontinued** for most participants:

- Only startups that reached **Level 4** in the former Founders Hub tier system may still qualify for **$5,000 in OpenAI credits**
- New participants should plan to use **Azure OpenAI Service** instead, which is a separate Microsoft product billed against your Azure credits
- Azure OpenAI provides access to GPT-5 mini/nano, GPT-4.1, GPT-4o, o4-mini, and other models directly within your Azure subscription. Advanced models (GPT-5.2, GPT-5.1, GPT-5, o3) require [limited-access registration](https://aka.ms/oai/gpt5access)

---

## Pegasus Program (Invite-Only Upgrade)

Top-performing startups in the investor offer may be invited to the **Pegasus Program**, which provides:

- Up to **$350,000 in total tech credits** across Azure, GitHub, and LinkedIn
- A **dedicated Cloud Solutions Architect** assigned to your startup
- Direct connections to **Microsoft enterprise customers** for sales opportunities
- Co-marketing and co-selling support

Pegasus is **invite-only** -- the Pegasus team monitors startups in the program and reaches out to those deemed a good fit based on solution area, maturity, traction, and team profile. You cannot apply directly.

---

## Credit Restrictions

Azure credits from this program **cannot** be used for:

- Non-Azure products
- Third-party products from Azure Marketplace
- Microsoft Support Plans (beyond the included Standard Support)

---

## What Happens When Credits Run Out

- Your Azure subscription converts to **pay-as-you-go** billing
- Running services are **not automatically stopped** -- you will be charged at standard rates
- Set up **budget alerts** in the Azure portal to monitor credit consumption
- Track remaining credits at the [Azure Sponsorship Portal](https://www.microsoftazuresponships.com/)

---

## Alternative: Basic Path (No Investor Required)

If you are not currently backed by an investor in the network, you can still access Azure credits through the **Azure Startup Credit Offer**:

- **$1,000** in Azure credits immediately (valid 90 days)
- **$4,000 additional** after business verification (valid 180 days)
- No referral code required
- Apply at [azure.microsoft.com/startups](https://learn.microsoft.com/en-us/azure/signups/overview)

Once you secure funding from a participating investor, you can then apply for the investor offer to access the larger credit pool.

---

## Additional Tips

- **The "Founders Hub" name is retired** as of July 2025 -- the program is now simply called "Microsoft for Startups"
- **Expert Network and Champion Network** sub-programs have also been retired
- **Credit activation clock** starts when you activate, not when you are approved -- do not activate until you are ready to use the credits
- **Region matters** for GPU availability -- check Azure region availability before planning AI workloads
- **Multiple investor codes** can be submitted to support for evaluation, but you receive one offer allocation
- **No equity taken** -- Microsoft does not take equity, require revenue share, or impose cohort participation

---

**Sources:**
- [Microsoft for Startups Investor Offer Benefits](https://learn.microsoft.com/en-us/microsoft-for-startups/benefits)
- [Microsoft for Startups Application Process](https://learn.microsoft.com/en-us/microsoft-for-startups/application)
- [Microsoft for Startups Overview](https://learn.microsoft.com/en-us/microsoft-for-startups/overview)
- [Changes to Microsoft for Startups - July 2025](https://learn.microsoft.com/en-us/microsoft-for-startups/changes-microsoft-for-startups)
- [Microsoft for Startups Investor Network FAQs](https://learn.microsoft.com/en-us/microsoft-for-startups/investor-faqs-for-learn-page)
- [Azure Credit Offer Details](https://learn.microsoft.com/en-us/microsoft-for-startups/benefits/azure)
- [Microsoft for Startups Portal](https://portal.startups.microsoft.com/)
- [Microsoft for Startups Pegasus Program](https://www.microsoft.com/en-us/startups/pegasus)
- [OpenAI Credits Discussion - OpenAI Community](https://community.openai.com/t/is-2-500-provided-by-microsoft-for-startups-founders-hub-still-a-valid-openai-offering/860544)

## Submission 48: Microsoft for Startups Founders Hub - Basic ($5,000)

- Category: Startup Programs
- Source Link: https://www.microsoft.com/en-us/startups
- Amount: $5,000 in Azure credits (starts at $1,000, unlocks to $5,000 after business verification)
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Microsoft for Startups Founders Hub gives early-stage founders **up to $5,000 in Azure credits** through a self-service path that requires no VC backing, no accelerator affiliation, and no equity. You start with **$1,000 for 90 days** just by signing in with LinkedIn, then unlock up to **$5,000 for 180 days** by verifying your business. On top of credits, you get **Microsoft 365 Business Premium (50 seats)**, **GitHub Enterprise (20 users)**, **Visual Studio Enterprise (5 users)**, and access to Azure technical experts. Credits work for Azure OpenAI Service (GPT-5 mini/nano, GPT-4.1, GPT-4o, o4-mini, and more), compute, databases, and most Azure services.

---

## Registration (Step-by-Step)

1. Go to [startups.microsoft.com](https://www.microsoft.com/en-us/startups)
2. Click **"Get started"** (or "Apply now")
3. **Sign in with LinkedIn** -- your LinkedIn profile is used for identity verification and program communications
4. Provide **contact details** -- use a business email matching your company domain (personal emails like @gmail.com have higher rejection rates)
5. Share your **founder background** and a brief description of your software product
6. Submit your application -- takes about **10 minutes** to complete
7. **Wait for review** -- decisions typically arrive via email within **3 business days**
8. Once accepted, you land at **Level 1** with **$1,000 in Azure credits** (valid 90 days)
9. To unlock the full **$5,000**, complete **business verification** within 90 days of initial redemption (see next section)

**Important:**
- You need a **LinkedIn account** to apply -- resolve any LinkedIn access issues before starting
- You **cannot edit** your application after submission. If you need to change something, delete and resubmit before a decision is made
- If declined, you must wait **14 days** before reapplying
- Activate your Azure benefits **within 3 months** of onboarding or you risk losing them

---

## Business Verification (Unlocking $5,000)

After receiving your initial $1,000 credit, you can verify your startup to unlock the remaining credits:

1. Go to the **Azure portal** and look for the verification prompt
2. Confirm your **business address**
3. Provide **business registration details** (incorporation documents)
4. Share a brief **startup description** explaining what your product does
5. Wait up to **7 business days** for processing
6. Check status via **"Check verification status"** on the Azure portal homepage

**Verification must be completed within 90 days** of redeeming your initial $1,000 credit, or you lose access to the additional credits.

---

## Credit Tiers & Progression

The Founders Hub has four levels. The basic (non-investor) path covers Levels 1-2:

| Level | Azure Credits | How to Unlock |
|---|---|---|
| **Level 1** | $1,000 | LinkedIn sign-in, application approved |
| **Level 2** | $5,000 | Business verification (registered entity) |
| **Level 3** | $25,000 | Domain verification + product demo |
| **Level 4** | $150,000 | Investor referral (VC/accelerator in Microsoft network) |

- Credits are **cumulative** -- Level 2 includes Level 1 credits already used. If you used $800 at Level 1 and unlock Level 2, you have $4,200 remaining
- Level 3 and Level 4 require progressively more validation and are part of the investor-affiliated path
- As your startup grows and Azure usage increases, you can request to move up levels

---

## Credit Expiration & Billing

| Credit Tier | Validity Period |
|---|---|
| **$1,000 (Level 1)** | 90 days from redemption |
| **$5,000 (Level 2)** | 180 days from verification approval |

- **No extensions available** -- unused credits expire automatically
- After credits expire or are exhausted, your subscription **auto-converts to pay-as-you-go**
- You **will be billed** for any usage after conversion unless you cancel or remove resources
- Set up **budget alerts** in the Azure portal immediately after activation
- Credits **do not roll over** between months or tiers

---

## What Credits Can and Cannot Be Used For

**Can use for:**
- Azure OpenAI Service (GPT-4o, GPT-4.1, GPT-4 Turbo, etc.)
- Azure VMs, Kubernetes, App Service
- Databases (Cosmos DB, PostgreSQL, SQL)
- Azure AI services (Speech, Vision, Language)
- Storage, networking, and most Azure services

**Cannot use for:**
- Microsoft Azure support plans
- Third-party branded products
- Azure Marketplace products
- Crypto mining or illegal activities
- Cannot be combined with other licensing, prepaid, or credit offers
- Cannot be transferred or shared between accounts

---

## Included Software & Tools

All accepted members receive the following at **no additional cost**:

| Product | Included |
|---|---|
| **Microsoft 365 Business Premium** | 50 seats |
| **GitHub Enterprise** | 20 users |
| **Visual Studio Enterprise** | 5 users |
| **Power Apps** | 10 users |
| **Power Automate** | 10 users |
| **Power BI** | 10 users |
| **Dynamics 365 Customer Service Enterprise** | 10 seats |
| **Dynamics 365 Sales Enterprise** | 10 seats |
| **Dynamics 365 Team Members** | 25 seats |
| **LinkedIn Premium** | Free |
| **Microsoft Clarity** | Free |

These are substantial -- GitHub Enterprise alone costs $21/user/month ($5,040/year for 20 users), and Visual Studio Enterprise is $250/month per user.

---

## Azure OpenAI & AI Services

Your Azure credits can be used for **Azure OpenAI Service**, which gives you access to:

- **GPT-5 mini**, GPT-5 nano, GPT-5 chat (generally available)
- **GPT-4.1**, GPT-4.1 mini, GPT-4.1 nano (1M token context)
- **GPT-4o**, GPT-4o mini
- **o4-mini**, **o3-mini** (reasoning models)
- **DALL-E**, **GPT Image 1** (image generation)

Azure OpenAI uses pay-per-token pricing charged against your credits. There is no separate free tier for Azure OpenAI -- it draws directly from your $5,000 credit balance.

**Note:** Advanced models (**GPT-5.2**, **GPT-5.1**, **GPT-5**, **o3**, **o3-pro**) require [limited-access registration](https://aka.ms/oai/gpt5access). Only Level 3+ members can request modifications to Azure OpenAI content filtering.

---

## Technical Support & Expert Access

Starting at Level 2 ($5,000 tier), you get:

- **Azure Standard Support** -- free 24/7 expert assistance (normally $100/month)
- **Azure Technical Experts** -- unlimited consultations with real engineers for architecture design, optimization, and scaling
- **Expert Network** -- unlimited access to mentors and advisors
- Guidance on deploying AI models, integrating with Azure services, and optimizing infrastructure costs

This is one of the most underrated benefits -- direct access to Azure architects who can review your infrastructure and suggest cost-saving approaches.

---

## Eligibility Requirements

**You qualify if:**
- You are building a **software-based product** (SaaS, AI tools, dev tools, platform, analytics)
- Your startup is a **registered legal business entity** (for Level 2+)
- Your company is headquartered in a **country where Azure is available**
- You have received **fewer than $350,000 in lifetime free Azure credits**
- You are **pre-Series C** funding stage
- You are a **privately held, for-profit business**

**You do NOT qualify if you are:**
- A consulting firm, development agency, or IT shop
- An educational institution or government organization
- Running a personal blog or non-software business
- Operating a crypto mining operation
- Licensing software from another party (must own your product)

**You do NOT need:**
- VC or accelerator backing
- Revenue or paying customers
- A launched product (concept stage is fine)
- A credit card (for the initial application -- only needed when activating Azure)

---

## Investor Path (Higher Credits)

If you have VC or accelerator backing from a firm in the **Microsoft for Startups Investor Network**, you can access higher credit tiers:

1. Request a **referral code** from your affiliated investor
2. Apply at [startups.microsoft.com](https://startups.microsoft.com/) using the referral code
3. Access up to **$150,000 in Azure credits** over four years
4. All software benefits are the same across both paths

This is worth pursuing if your investor is in the network -- ask them directly.

---

## Additional Tips

- **Use a business email** matching your company domain when applying. Personal emails (@gmail, @yahoo) significantly reduce your chances of approval
- **The credit clock starts immediately** upon redemption -- do not activate until you are ready to build
- **Budget alerts are essential** -- set them up in the Azure portal on day one to avoid surprise charges after credits expire
- **Compare with Azure Free Account** -- new Azure customers can also get a separate $200 free trial. These programs cannot be combined, so choose the one that fits your needs
- **Third-party benefits change** -- the Founders Hub used to include direct OpenAI API credits ($2,500), but this has been discontinued. Use Azure OpenAI Service instead
- **Region matters** -- not all Azure OpenAI models are available in all regions. Check [regional availability](https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models) before creating resources
- **Level progression takes time** -- do not expect to jump from Level 1 to Level 3 quickly. Build traction and demonstrate Azure usage growth to justify level-up requests

---

**Sources:**
- [Microsoft for Startups](https://www.microsoft.com/en-us/startups)
- [Azure Startup Credits Overview](https://learn.microsoft.com/en-us/azure/signups/overview)
- [Founders Hub Application Process](https://learn.microsoft.com/en-us/microsoft-for-startups/application)
- [Founders Hub Benefits Guide](https://partner.microsoft.com/en-us/partnership/founders-hub-benefits-guide)
- [Investor Offer Benefits](https://learn.microsoft.com/en-us/microsoft-for-startups/benefits)
- [Azure OpenAI for Startups Blog](https://www.microsoft.com/en-us/startups/blog/azure-openai-service-for-startups/)
- [Founders Hub Benefits Blog](https://www.microsoft.com/en-us/startups/blog/founders-hub-benefits/)

## Submission 47: Microsoft Imagine Cup - Azure Credits for Student Competitors

- Category: Education & Student
- Source Link: https://imaginecup.microsoft.com/
- Amount: $1,000 in Azure credits + $2,500 in OpenAI credits (qualifying teams); up to $5,000 Azure credits for semifinalists/MVPs; cash prizes up to $100,000
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Student teams competing in Microsoft's annual **Imagine Cup** competition receive **$1,000 in Azure credits** and **$2,500 in OpenAI credits** through the Microsoft for Startups Founders Hub -- with no credit card required. Semifinalists unlock an additional **$4,000 in Azure credits** (Founders Hub Level 2). MVPs can receive up to **$5,000** total. The competition runs annually from late September through the World Championship at Microsoft Build, with cash prizes of **$100,000** (Scale winner) and **$50,000** (Launch winner).

---

## Registration & Credits (Step-by-Step)

1. Go to [imaginecup.microsoft.com](https://imaginecup.microsoft.com) and click **"Register Now"**
2. Sign in with a **personal Microsoft account** (Outlook, Hotmail, or any personal Microsoft-linked email). **Do not use a school or work account** -- it will cause issues with Founders Hub enrollment
3. Create your team (1-4 members). The team leader adds other members by email through the dashboard. Members can be from different institutions and different countries
4. Submit your AI startup idea via the dashboard under **"Submit Idea"** or **"MVP Round Entry"**. Upload a PDF Lean Canvas, a pitch deck, a recorded pitch video, and a demo video
5. After submission, go to [portal.startups.microsoft.com/signup](https://portal.startups.microsoft.com/signup) and apply to **Microsoft for Startups Founders Hub**
6. During the Founders Hub application, select **"Imagine Cup"** as your partner affiliation
7. Once accepted, your **$1,000 Azure credits** and **$2,500 OpenAI credits** will appear in your Azure portal (allow 1-3 business days for credits to show up)

**Important:**
- Use the **same personal Microsoft account** for both Imagine Cup registration and Founders Hub
- Credits must be **activated within 90 days** of being awarded
- No credit card is required at any step
- If credits do not appear after 3 days, file a support ticket through the Imagine Cup portal with proof of registration

---

## Credit Breakdown by Stage

| Competition Stage | Azure Credits | OpenAI Credits | Founders Hub Level |
|---|---|---|---|
| **Qualifying (all accepted teams)** | $1,000 USD | $2,500 USD | Level 1 |
| **Semifinalists** | +$4,000 USD (total $5,000) | Included | Level 2 |
| **MVPs** | Up to $5,000 USD | Included | Level 2 |

Credits are delivered through Microsoft for Startups Founders Hub and can be used for **any Azure service**, including Azure OpenAI, Azure AI Foundry, compute, databases, and storage.

---

## Microsoft AI Services Requirement

Your submission must use **at least two (2) Microsoft AI services** to operate. Qualifying services include:

- **Azure OpenAI Service** (GPT-4o, GPT-4.1, etc.)
- **Azure AI Foundry** (model deployment and management)
- **Azure Machine Learning**
- **Azure AI Search**
- **Azure AI Vision / Computer Vision**
- **Azure AI Language** (sentiment analysis, NER, key phrases)
- **Azure AI Speech** (speech-to-text, text-to-speech)
- **Azure AI Document Intelligence**
- **Azure AI Content Safety**

Using GitHub Copilot as a development tool does **not** count toward the two-service requirement -- the services must be integrated into your product.

---

## Competition Paths: Scale vs Launch

| | **Scale** | **Launch** |
|---|---|---|
| **For** | Ventures with existing traction, users, or revenue | Teams building foundations, testing prototypes, refining MVPs |
| **Funding cap** | Up to $5M in combined dilutive or non-dilutive funding received | No dilutive funding, max $100K in external non-dilutive funding |
| **Top prize** | $100,000 USD + mentorship with Satya Nadella | $50,000 USD |
| **World Championship** | Top 3 teams advance | Winner selected at semifinals |

Microsoft reserves the right to reassign teams between paths before April 7, 2026.

---

## Competition Timeline (2026 Season)

| Phase | Dates |
|---|---|
| **Registration opens** | September 30, 2025 |
| **Idea / MVP submission deadline** | January 9, 2026 |
| **MVP Submissions** | January 26 -- February 9, 2026 |
| **Semifinals** | February 23 -- April 12, 2026 |
| **World Championship** | April 26, 2026 -- Microsoft Build |
| **Winners announced** | By June 30, 2026 |

Timelines are annual and shift slightly each year. Check the [Imagine Cup website](https://imaginecup.microsoft.com) for the current season's exact dates.

---

## Eligibility Requirements

- **Age:** Must be at least **18 years old** as of September 30, 2025
- **Student status:** Must be enrolled at an **accredited educational institution** (high school or college/university) at any time between September 30, 2025 and May 31, 2026
- **Team size:** 1 to 4 eligible students per team
- **Cross-institution:** Team members can be from different schools and different countries
- **Country:** Must be from a country not restricted by U.S. export regulations
- **Graduates** who are no longer enrolled do **not** qualify
- **MOOCs and for-profit training programs** do not count as accredited institutions

---

## Submission Requirements

Each team must submit:

1. **Pitch deck** (PDF) -- Lean Canvas format covering your AI solution and Azure AI integration
2. **Recorded pitch video** -- explaining your product vision and market opportunity
3. **Recorded demo video** -- showing the working product built with Microsoft AI services
4. All submissions must address **diversity, inclusion, and accessibility** considerations

---

## Judging Criteria

| Criterion | Weight |
|---|---|
| **Viability** (new category creation, clear need, credible market plan) | 50% |
| **AI Technology** (AI integration, technical design, responsible AI) | 40% |
| **Inclusion** (accessibility and inclusivity for all users) | 10% |

---

## Competition Categories

Teams are encouraged (but not required) to align with one of these focus areas:

- Cybersecurity
- Education
- Financial Services
- Health and Life Sciences
- Retail and Consumer Goods
- Self-defined custom category

Solutions outside these categories may still be considered at Microsoft's discretion.

---

## Additional Tips

- **Start with Founders Hub early** -- do not wait until the submission deadline to apply. Credit activation can take several days, and you need Azure resources to build your MVP
- **IP stays with you** -- Microsoft does not claim ownership of any submissions or intellectual property
- **GitHub Student Developer Pack** is available to all student participants and includes GitHub Copilot, free domain names, cloud credits from other providers, and development tools
- **Replit access** is available for semifinalists to help with rapid prototyping
- **Mentorship** -- semifinalists receive personalized mentorship on architecture, go-to-market strategy, and AI acceleration
- **Related programs:** If you do not make it into Imagine Cup, consider [Azure for Students](https://azure.microsoft.com/en-us/free/students) ($100 credit, no credit card, renewable annually) or the general [Microsoft for Startups Founders Hub](https://www.microsoft.com/en-us/startups) (up to $5,000 in Azure credits for any startup founder)
- **Past winners** include teams building solutions for accessibility, healthcare diagnostics, and agricultural technology -- practical impact is valued over flashy demos

---

**Sources:**
- [Imagine Cup Official Website](https://imaginecup.microsoft.com)
- [Imagine Cup FAQ](https://imaginecup.microsoft.com/en-us/support/faq)
- [Imagine Cup 2026 Official Rules & Regulations (PDF)](https://imaginemediastoragemiauth-h6a2djhgcddwgkd7.z01.azurefd.net/content/IC26%20Official%20Rules%20and%20Regulations%20Dec%202025.pdf)
- [Imagine Cup 2026 Blog Post -- Microsoft Tech Community](https://techcommunity.microsoft.com/blog/studentdeveloperblog/imagine-cup-2026-where-students-build-what%E2%80%99s-next/4456700)
- [Azure Credits for Imagine Cup -- Microsoft Q&A](https://learn.microsoft.com/en-us/answers/questions/5674199/how-can-i-access-free-azure-credits-for-imaginatio)
- [Microsoft for Startups Founders Hub](https://www.microsoft.com/en-us/startups)
- [Azure for Students](https://azure.microsoft.com/en-us/free/students)

## Submission 46: Lightning AI Free Plan - 22 GPU Hours/Month

- Category: Free GPU & Compute
- Source Link: https://lightning.ai/pricing
- Amount: 15 monthly credits (~22 GPU hours/month on T4); 10 GB free Drive storage; 100 GB Studio storage
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Lightning AI offers a **free cloud-based AI development platform** with access to modern NVIDIA GPUs (T4, L4, A10G, L40S). Free users receive **15 monthly Lightning credits** (each credit is worth approximately $1), which translates to roughly **22 hours of T4 GPU time per month**. The platform provides a browser-based IDE similar to VS Code, with no local setup required. No credit card is needed to sign up.

---

## Registration (Step-by-Step)

1. Go to [lightning.ai/sign-up](https://lightning.ai/sign-up)
2. Create an account using email, Google, or GitHub
3. Use an **official company or .edu email** for instant verification — personal emails may require additional steps
4. **Verify your phone number** to unlock an initial bonus of 7 free GPU hours
5. Follow the onboarding steps to customize your Studio
6. Done — you will be placed in a sample project with a basic Python file ready to run

**Important:**
- No credit card required for the free plan
- Some users on Lightning AI forums have reported delays in receiving free credits — if this happens, contact Lightning AI support
- Credits renew monthly; unused credits do **not** roll over

---

## GPU Options & Credit Costs

Lightning credits are consumed based on the GPU type you select. Each credit is worth approximately $1. Here are the approximate consumption rates:

| GPU | VRAM | Credits/Hour | Hours per 15 Credits |
|---|---|---|---|
| **T4** | 16 GB | ~0.68 | ~22 hours |
| **L4** | 24 GB | ~0.70 | ~21.4 hours |
| **A10G** | 24 GB | ~1.80 | ~8.3 hours |
| **L40S** | 48 GB | ~2.00 (est.) | ~7.5 hours (est.) |

**Free plan users are limited to single-GPU Studios only.** Multi-GPU configurations require the Pro plan or higher.

To maximize your free credits, use the **T4 or L4** GPUs for training and inference tasks. Reserve the A10G or L40S for workloads that truly need more VRAM or compute.

---

## Studio Features

### Included in the Free Plan
- **1 free active Studio** with 4-hour automatic restarts (the Studio shuts down and restarts every 4 hours; your data persists)
- **Browser-based IDE** (VS Code-like interface) or connect your **local IDE** (VS Code, PyCharm) via SSH
- **Jupyter Notebook** support — switch between VS Code and Jupyter interfaces
- **32-core CPU Studios** available for CPU-only workloads
- **Unlimited background execution** — jobs keep running even when the browser is closed
- **Pre-configured ML environments** — PyTorch, TensorFlow, and common ML libraries are pre-installed
- **Multiplayer collaboration** — invite others to code together in real-time
- **SDK automation** — programmatically manage Studios via the Lightning SDK
- **Community Templates** — start from pre-built Studios for common tasks (fine-tuning, training, deployment)
- **App hosting** — deploy and share AI apps built with Streamlit, Gradio, or React

### Free Plan Restrictions
- **4-hour restart cycle** — your Studio automatically restarts every 4 hours (Pro plan gets 24/7 uptime)
- **Single GPU only** — no multi-GPU training
- **No A100, H100, or H200 GPUs** — these require the Teams plan ($140/user/month)
- **100 GB storage limit** on Studios
- **Community support only** (Discord) — no priority support

---

## Storage

| Storage Type | Free Limit | Notes |
|---|---|---|
| **Studio storage** | 100 GB | Persistent across restarts; files survive the 4-hour restart cycle |
| **Drive storage** | 10 GB | Shared file storage; upload/download files; connect S3 buckets |

Studio storage persists between sessions, so you will not lose your work when the 4-hour restart occurs. Files, models, and datasets remain intact.

---

## Earning Extra Credits

Lightning AI offers a **quests system** where users can complete tasks to earn bonus credits — reportedly up to 240 additional credits. Quests typically involve exploring platform features. Check the Lightning AI dashboard for available quests after signing up.

---

## Paid Plan Comparison

| Feature | Free | Pro ($50/mo annual) | Teams ($140/user/mo annual) |
|---|---|---|---|
| Monthly credits | 15 | 40 | 50 |
| Studio uptime | 4-hour restarts | 24/7 | 24/7 |
| GPU types | T4, L4, A10G, L40S (single) | T4, L4, A10G, L40S (multi) | + A100, H100, H200 |
| Storage | 100 GB | 2 TB | Unlimited |
| Spot instances | No | No | Yes (~80% savings) |
| Multi-node training | No | No | Up to 4 machines |

---

## Additional Tips

- **Use T4 GPUs to stretch your credits** — at ~0.68 credits/hour, you get the most compute time from your 15 free credits
- **Switch to CPU when not training** — coding, debugging, and data prep do not need a GPU. Use the free 4-CPU Studio for these tasks to save credits
- **Phone verification is essential** — without it, you may not receive your initial 7 bonus GPU hours
- **The 4-hour restart does not delete your files** — storage is persistent, so you can safely leave long-running jobs with checkpointing
- **Use community templates** — pre-built Studios save setup time and let you jump straight into fine-tuning or training
- **Lightning AI CI/CD runners** — you can use Lightning AI as free GPU runners for CI/CD pipelines (e.g., for automated ML testing)
- **Compare with Google Colab** — Lightning AI offers a more IDE-like experience with persistent storage and better GPU options compared to Colab's free tier, though Colab does not require phone verification

---

**Sources:**
- [Lightning AI Pricing](https://lightning.ai/pricing/)
- [Lightning AI Billing FAQ](https://lightning.ai/docs/overview/faq/billing)
- [Lightning AI Consumption Rates](https://lightning.ai/pricing/consumption-rates)
- [Lightning AI Billing Documentation](https://lightning.ai/docs/overview/billing)
- [Lightning AI Sign Up](https://lightning.ai/sign-up)
- [KDnuggets: Using Lightning AI Studio For Free](https://www.kdnuggets.com/using-lightning-ai-studio-for-free)
- [GMI Cloud: Free GPU Cloud Trials Guide](https://www.gmicloud.ai/blog/where-can-i-get-free-gpu-cloud-trials-in-2025-a-complete-guide)

## Submission 45: Lepton AI / NVIDIA DGX Cloud Lepton - Free Plan

- Category: AI API Free Tiers
- Source Link: https://www.lepton.ai/
- Amount: Free plan with up to 48 CPUs + 2 GPUs concurrently, 1 GB storage free, 10 GB network egress free, serverless endpoints at pay-per-token rates
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Lepton AI (now **NVIDIA DGX Cloud Lepton**) offers a free Basic Plan with **no subscription fees** and pay-as-you-go pricing. The free plan includes up to **48 CPUs and 2 GPUs concurrently** in a single workspace. The platform provides serverless endpoints for popular open-source LLMs with **OpenAI-compatible API**, dedicated GPU compute billed by the minute, and a Pythonic SDK for building and deploying custom AI services. NVIDIA acquired Lepton AI in April 2025 and rebranded it as DGX Cloud Lepton, connecting developers to tens of thousands of GPUs from a global network of cloud providers.

---

## Registration (Step-by-Step)

1. Go to [lepton.ai](https://www.lepton.ai/) (redirects to nvidia.com/en-us/data-center/dgx-cloud-lepton/)
2. Click **"Get Started"** or **"Sign Up"**
3. Create an account using your email or a social login (GitHub, Google)
4. After signing in, access the **Lepton Dashboard** to create a workspace
5. Install the Python SDK locally: `pip install -U leptonai`
6. Authenticate your CLI with: `lep login` and follow the prompts to link your account
7. You are now on the **Basic Plan** with no subscription fee — you only pay for resources you actually consume

**Important:**
- The Basic Plan is a single-user workspace
- Rate limit for serverless API endpoints on Basic Plan is **10 requests per minute**
- No upfront payment or subscription required — purely usage-based billing
- The platform now redirects to NVIDIA's domain, but the Lepton dashboard and docs remain functional

---

## Free Plan Limits (Basic Plan)

| Resource | Limit |
|---|---|
| **Subscription fee** | $0/month |
| **Max concurrent CPUs** | 48 |
| **Max concurrent GPUs** | 2 |
| **Workspace users** | 1 |
| **Serverless API rate limit** | 10 RPM |
| **Storage** | First 1 GB free ($0.153/GB/month after) |
| **Network egress** | First 10 GB/month free ($0.15/GB after) |

---

## Available Serverless Models

Lepton AI provides pre-configured serverless endpoints for popular open-source models. All LLM endpoints are **fully compatible with OpenAI's API spec** — you can use them as a drop-in replacement with OpenAI client libraries.

Since the NVIDIA acquisition, Lepton has shifted from a fixed model catalog to an **infrastructure-first platform** connecting developers to 10K+ concurrent models across 20+ NVIDIA Cloud Partners. The previous fixed serverless model lineup (Llama 3.x, Mixtral, Mistral 7B) has been superseded by a dynamic model ecosystem.

The platform now focuses on:
- **Custom model deployment** — bring your own models via the Pythonic SDK
- **Pre-built templates** — deploy popular open-source models (Llama 4, Qwen3, DeepSeek, etc.) from Hugging Face
- **Partner model access** — inference through cloud partners (Together AI, Nebius, Scaleway, etc.)

**Note:** Specific model availability and pricing are dynamic and depend on the cloud partner routing. Check the [Lepton dashboard](https://dashboard.lepton.ai/) for current serverless endpoints and pricing.

Beyond LLMs, Lepton also offers serverless endpoints for:
- **Whisper** — audio transcription
- **Stable Diffusion** — image generation

---

## Dedicated GPU Compute Pricing

For custom deployments, training, or running your own models, Lepton offers dedicated GPU instances billed by the minute:

| GPU Type | vRAM | RAM | vCPU | Price/min | ~Price/hr |
|---|---|---|---|---|---|
| **NVIDIA A10** | 24 GB | 96 GB | 24 | $0.0202 | ~$1.21 |
| **NVIDIA RTX A6000** | 48 GB | 64 GB | 8 | $0.0275 | ~$1.65 |
| **NVIDIA H100 80GB** | 80 GB | 240 GB | 20 | $0.0500 | ~$3.00 |
| **NVIDIA A100 80GB** | 80 GB | 192 GB | 12 | $0.0535 | ~$3.21 |

- A100 and H100 instances can scale to **1, 2, 4, or 8 GPUs**
- Compute is billed by the minute with no minimum commitment
- On the Basic Plan, you can use up to **2 GPUs concurrently**

---

## Platform Features

- **OpenAI-compatible API** — use the standard OpenAI Python/JS client with Lepton's base URL
- **Pythonic SDK (leptonai)** — build models with Python, deploy with `lep` CLI, no Docker/Kubernetes needed
- **Dev Pods** — interactive development via Jupyter notebooks, SSH, or VS Code
- **Batch Jobs** — large-scale training and data preprocessing across multiple nodes
- **Custom Model Deployment** — deploy any model from HuggingFace or your own code as a service ("Photon")
- **Auto-scaling** — serverless endpoints scale automatically based on traffic
- **Multi-cloud** — since the NVIDIA acquisition, the platform connects to 20+ cloud providers (CoreWeave, Lambda, Nebius, Crusoe, AWS, and more)

---

## NVIDIA DGX Cloud Lepton Transition

NVIDIA acquired Lepton AI in early April 2025 for several hundred million dollars. Key changes:

- **Rebranded** from "Lepton AI" to **"NVIDIA DGX Cloud Lepton"**
- **lepton.ai** now redirects to **nvidia.com/en-us/data-center/dgx-cloud-lepton/**
- The legacy Lepton Dashboard remains accessible for existing customers
- Co-founders **Yangqing Jia** (ex-Alibaba VP, creator of Caffe) and **Junjie Bai** stayed on post-acquisition
- The platform is being expanded as a **GPU compute marketplace** connecting 20+ NVIDIA Cloud Partners, including access to NVIDIA Blackwell GPUs
- Future integration with **NVIDIA NIM microservices**, **NeMo**, and **Cloud Functions (NVCF)**

**What this means for free plan users:** The Basic Plan still exists as of early 2026, but the platform is in transition. Pricing, available models, and plan structures may change as NVIDIA integrates Lepton into DGX Cloud.

---

## Paid Plans

| Plan | Monthly Fee | Max CPUs | Max GPUs | Serverless RPM | Users |
|---|---|---|---|---|---|
| **Basic (Free)** | $0 | 48 | 2 | 10 | 1 |
| **Standard** | $30 | 192 | 16 | 600 | Multi-user |
| **Enterprise** | Custom | Custom | Custom | Custom | Unlimited |

The Standard plan adds multi-user workspaces, higher concurrency, dedicated support, and advanced features. Enterprise adds custom SLAs, account management, and 24/7 priority support.

---

## Additional Tips

- **No auto-billing surprise** — the Basic Plan has no subscription fee. You only pay for compute, storage, and network you actually use. If you do not spin up any resources, you pay nothing
- **OpenAI SDK compatibility** — to use Lepton's serverless endpoints, just change the `base_url` in your OpenAI client to `https://<model-name>.lepton.run/api/v1/` and use your Lepton API key
- **Storage under 1 GB is free** — ideal for small experiments and prototypes
- **10 RPM is limiting** — the Basic Plan's 10 requests per minute rate limit on serverless endpoints is tight for production use. It is best for experimentation and prototyping
- **GPU availability may vary** — since DGX Cloud Lepton aggregates across cloud providers, GPU availability depends on the partner network and current demand
- **Python 3.10+ recommended** — the leptonai SDK works best with Python 3.10 or newer
- **Platform is in transition** — with the NVIDIA acquisition still relatively recent, expect features, pricing, and branding to evolve. The legacy Lepton dashboard is still available but may eventually migrate fully to NVIDIA's infrastructure

---

**Sources:**
- [NVIDIA DGX Cloud Lepton (lepton.ai)](https://www.lepton.ai/)
- [Lepton AI Pricing](https://www.lepton.ai/pricing)
- [Lepton AI Billing and Payments](https://www.lepton.ai/docs/platform/billing_n_payments)
- [Lepton AI Quickstart Guide](https://www.lepton.ai/docs/guides/get-started/quickstart)
- [Lepton AI Serverless Endpoints Documentation](https://www.lepton.ai/docs/guides/endpoints/serverless-endpoints)
- [Introducing NVIDIA DGX Cloud Lepton (NVIDIA Developer Blog)](https://developer.nvidia.com/blog/introducing-nvidia-dgx-cloud-lepton-a-unified-ai-platform-built-for-developers/)
- [NVIDIA Announces DGX Cloud Lepton (NVIDIA Newsroom)](https://nvidianews.nvidia.com/news/nvidia-announces-dgx-cloud-lepton-to-connect-developers-to-nvidias-global-compute-ecosystem)
- [Lepton AI on Artificial Analysis](https://artificialanalysis.ai/providers/leptonai)
- [Lepton AI GitHub](https://github.com/leptonai/leptonai)

## Submission 44: Lambda Labs Research Grant - Up to $5,000

- Category: Free GPU & Compute
- Source Link: https://lambda.ai/research
- Amount: Up to $5,000 in cloud credits
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Lambda Labs offers **up to $5,000 in cloud credits** to qualifying academic researchers through its Research Grant program. There is **no general-purpose free trial or free tier** for regular users — this is a competitive, application-based grant. Recipients get access to Lambda's GPU cloud instances (NVIDIA B200, H100, A100, A6000, and more) plus mentoring from Lambda's Chief Scientific Officer, Chuan Li. Lambda expects to sponsor hundreds of researchers, with funded work presented at top venues like NeurIPS, ICML, and ICCV.

---

## Who Is This For?

This program is specifically for **academic researchers** working on AI/ML projects. It is not for indie hackers, hobbyists, or startup founders looking for general compute credits. You need:

- An active research project in AI/ML (focus areas include multimodal AI, generative AI, reasoning, and scaling)
- Affiliation with a research institution or university
- A project compelling enough to pass Lambda's review process

If you are not an academic researcher, Lambda Cloud has no free trial or signup credits. You will need to pay on-demand rates starting at **$0.50/hr** (Quadro RTX 6000) and up.

---

## How to Apply for the Research Grant

1. Go to [lambda.ai/research](https://lambda.ai/research)
2. Review the featured research projects to understand the caliber of work Lambda funds
3. Click the **Apply** button (links to a Typeform application)
4. Fill out the application with your research proposal, institutional affiliation, and compute requirements
5. Submit and wait for review — there is **no publicly stated timeline** for decisions
6. If accepted, credits are added to your Lambda Cloud account

**Important:**
- There is no published application deadline — the program appears to accept applications on a rolling basis
- There is no guarantee of receiving the full $5,000; the grant is described as "up to $5,000"
- Select research may be featured on Lambda's website
- You also receive mentoring from Chuan Li, Lambda's Chief Scientific Officer

---

## Lambda Cloud Account Setup (Required Before or After Grant Approval)

1. Go to [cloud.lambdalabs.com/sign-up](https://cloud.lambdalabs.com/sign-up)
2. Create an account with your email address
3. **Phone verification** — provide a phone number for SMS verification
4. **Credit card required** — you must register a valid payment method even if you have grant credits
5. **Add an SSH key** — upload your public SSH key (supports OpenSSH, RFC4716, PKCS8, PEM formats)
6. Launch an instance from the dashboard — select GPU type, region, and attach persistent storage

**Note:** Credit card registration is mandatory. Grant credits are applied as service credits that decrease at the end of each weekly billing cycle.

---

## Available GPU Instances

Grant credits can be used on any Lambda Cloud on-demand instance. Here is the full pricing table (prices are per GPU per hour):

### 1x GPU Instances

| GPU | VRAM | vCPUs | RAM | Storage | $/hr |
|---|---|---|---|---|---|
| **B200 SXM6** | 180 GB | 26 | 360 GB | 2.75 TB SSD | $5.29 |
| **GH200** | 96 GB | 64 | 432 GB | 4 TB SSD | $1.99 |
| **H100 SXM** | 80 GB | 26 | 225 GB | 2.75 TB SSD | $3.29 |
| **H100 PCIe** | 80 GB | 26 | 225 GB | 1 TB SSD | $2.49 |
| **A100 SXM** | 40 GB | 30 | 220 GB | 512 GB SSD | $1.29 |
| **A100 PCIe** | 40 GB | 30 | 225 GB | 512 GB SSD | $1.29 |
| **A10** | 24 GB | 30 | 226 GB | 1.3 TB SSD | $0.75 |
| **A6000** | 48 GB | 14 | 100 GB | 512 GB SSD | $0.80 |
| **Quadro RTX 6000** | 24 GB | 14 | 46 GB | 512 GB SSD | $0.50 |

### Multi-GPU Instances (selected)

| GPU | Config | $/hr per GPU |
|---|---|---|
| **B200 SXM6** | 8x | $4.99 |
| **H100 SXM** | 8x | $2.99 |
| **A100 SXM 80 GB** | 8x | $1.79 |
| **A100 SXM 40 GB** | 8x | $1.29 |
| **Tesla V100** | 8x | $0.55 |
| **A6000** | 4x / 2x | $0.80 |
| **A100 PCIe 40 GB** | 4x / 2x | $1.29 |

### What $5,000 Gets You (Approximate Hours)

| GPU | Approx. Hours on $5,000 |
|---|---|
| Quadro RTX 6000 (1x) | ~10,000 hrs |
| A6000 (1x) | ~6,250 hrs |
| A10 (1x) | ~6,667 hrs |
| A100 PCIe 40 GB (1x) | ~3,876 hrs |
| H100 PCIe (1x) | ~2,008 hrs |
| H100 SXM (1x) | ~1,520 hrs |
| B200 SXM6 (1x) | ~945 hrs |

---

## Software Stack

Every Lambda instance comes pre-installed with **Lambda Stack**:

- NVIDIA drivers and **CUDA toolkit**
- **cuDNN**, **NCCL**, **OFED** (for InfiniBand networking)
- **PyTorch**, **TensorFlow**, **JAX**
- **Docker** and **JupyterLab**
- Ubuntu-based OS

A minimal "GPU Base" image (Ubuntu Server with drivers only) is also available.

---

## Persistent Storage & Data

- Create **persistent storage filesystems** that mount at `/lambda/nfs/<FILESYSTEM_NAME>`
- Filesystems must be attached **before** launching an instance
- Data on terminated instances is **permanently lost** unless saved to persistent storage
- **No egress fees** — download your data, checkpoints, and model weights without additional charges

---

## Regions & Networking

- Instances are available in **US regions** (Utah us-west-3, Texas us-south-1)
- Network bandwidth: up to **200 Gbps** between instances, **20 Gbps** to the internet
- InfiniBand networking is included as standard for multi-GPU and cluster workloads
- SOC 2 Type II compliant, single-tenant infrastructure

---

## Capacity & Availability

Lambda Cloud operates on a **first-come, first-served** basis with no reservations for on-demand instances. Popular GPU types (especially H100 and B200) frequently experience **capacity shortages**. This is a known issue — if your preferred GPU is not available, you may need to check back repeatedly or consider a less popular configuration.

---

## Additional Tips

- **No free trial exists** — Lambda does not offer signup credits or a free tier for general users. The research grant is the only way to get free compute
- **Reserved pricing** — if you commit to 1-3 year terms, Lambda offers discounts up to 50% off on-demand rates (e.g., H100 cluster at $2.19/GPU-hr reserved vs $2.29 on-demand)
- **1-Click Clusters** — for large-scale training, Lambda offers clusters of 16-2,000+ interconnected B200 or H100 GPUs
- **Billing is weekly** — charges and credit deductions happen at the end of each weekly billing cycle, not in real-time
- **Google for Startups, AWS, Azure** — if you are not an academic researcher, consider free credit programs from major cloud providers instead (GCP $300, Azure $200, AWS research credits)
- **Lambda for Startups** — Lambda may have startup-specific programs; contact their sales team directly

---

**Sources:**
- [Lambda Research Grant Program](https://lambda.ai/research)
- [Lambda Cloud Pricing](https://lambda.ai/pricing)
- [Lambda Cloud Instances](https://lambda.ai/instances)
- [Lambda Cloud Getting Started Guide](https://docs.lambda.ai/public-cloud/on-demand/getting-started/)
- [Lambda Cloud Sign Up](https://cloud.lambdalabs.com/sign-up)
- [Lambda Research Grant Announcement (BusinessWire)](https://www.businesswire.com/news/home/20241209358457/en/Lambda-Launches-Grant-for-AI-Researchers-Expands-Research-Program)
- [Lambda Research Grant Announcement (HPCwire)](https://www.hpcwire.com/off-the-wire/lambda-launches-grant-for-ai-researchers-expands-research-program/)
- [Free GPU Cloud Trials Guide (GMI Cloud)](https://www.gmicloud.ai/blog/where-can-i-get-free-gpu-cloud-trials-in-2025-a-complete-guide)

## Submission 43: Kaggle Free GPU & TPU - 30 Hours/Week

- Category: Free GPU & Compute
- Source Link: https://www.kaggle.com/code
- Amount: 30 hours/week GPU + 20–30 hours/week TPU (floating quota)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Kaggle (owned by Google) gives every verified account holder **free weekly access to NVIDIA GPUs and Google TPU v3-8 accelerators** through browser-based Jupyter notebooks. No credit card is required. You get **30 hours/week of GPU time** and a **separate 20–30 hours/week of TPU time** (Kaggle uses a floating quota system that can vary with demand). Sessions run up to 9 hours for GPU/TPU notebooks and up to 12 hours for CPU-only notebooks, with background execution so training continues after you close the browser tab.

---

## Registration (Step-by-Step)

1. Go to [kaggle.com](https://www.kaggle.com) and click **Register**
2. Sign up with a **Google account**, email, or other supported method — completely free
3. Navigate to your **profile settings** (click your avatar → Settings)
4. Scroll to **Phone Verification** and verify your phone number via SMS code
5. Phone verification is **required** to unlock GPU and TPU accelerators — without it you only get CPU notebooks
6. Once verified, open or create any notebook at [kaggle.com/code](https://www.kaggle.com/code), click **Settings** in the right sidebar, and select your accelerator under the **Accelerator** dropdown (GPU P100, GPU T4 x2, or TPU v3-8)

**Important:**
- One phone number per account — Kaggle enforces this to prevent abuse
- Some users report difficulties with phone verification (certain carriers or VoIP numbers may not work)
- No billing account or credit card is ever required
- You can start using CPU notebooks immediately without phone verification

---

## GPU & TPU Specs

| Accelerator | VRAM / Memory | Details |
|---|---|---|
| **NVIDIA Tesla P100** | 16 GB HBM2 | 3,584 CUDA cores; strong FP32 performance; best for standard precision training |
| **NVIDIA T4 x2 (beta)** | 2 × 16 GB GDDR6 (32 GB total) | Tensor Cores optimized for FP16 / mixed precision; use `DataParallel` for distributed training |
| **TPU v3-8** | 128 GB HBM (8 cores) | Google's custom ASIC; excellent for large-batch TensorFlow and JAX workloads |

### Choosing Between P100 and T4 x2

- **P100** is faster for pure FP32 workloads — roughly 1.5× faster than dual T4 in FP32 due to superior memory bandwidth
- **T4 x2** shines with **mixed precision (AMP / FP16)** training — Tensor Cores go unused without AMP enabled
- Dual T4 can suffer from CPU bottlenecks (CPU must decode data for both GPUs) and inter-GPU communication overhead
- For most deep learning tasks with AMP enabled, T4 x2 offers more total VRAM (32 GB vs 16 GB)

---

## Weekly Quotas & Session Limits

| Resource | Limit |
|---|---|
| **GPU quota** | 30 hours/week (shared across P100 and T4 x2) |
| **TPU quota** | 20–30 hours/week (floating quota, varies with demand) |
| **GPU/TPU session max** | 9 hours per session |
| **CPU-only session max** | 12 hours per session |
| **CPU weekly quota** | Unlimited (no weekly cap, only per-session limit) |
| **Quota reset** | Rolling weekly window |

**Notes:**
- Kaggle uses a "floating" quota system — your weekly hours may vary slightly based on overall platform demand
- If you exhaust your GPU quota mid-week, you can still run CPU-only notebooks without restriction
- Interactive sessions (in the browser editor) will prompt "Are you still there?" after inactivity and may terminate if not confirmed

---

## Background Execution

Kaggle supports **background execution** via the "Save & Run All (Commit)" feature:

1. Click **Save Version** → select **Save & Run All (Commit)**
2. A separate background kernel is launched that runs independently of your browser
3. You can close the tab, and the notebook continues executing in the background
4. Output files are saved to `/kaggle/working/` and become downloadable once the run completes
5. The same session time limits apply (9 hours for GPU/TPU, 12 hours for CPU)

This is the recommended approach for long training runs — interactive sessions will time out if you walk away.

---

## Resources Per Session

| Resource | Amount |
|---|---|
| **CPU cores** | 4 vCPUs |
| **RAM (GPU notebooks)** | ~29 GB |
| **RAM (CPU notebooks)** | ~16 GB |
| **Disk (working directory)** | 20 GB (`/kaggle/working/`) |
| **Output file limit** | 500 files max in output directory |
| **Internet access** | Available (must be enabled in Settings; disabled by default in some competitions) |

---

## Storage & Data

- **/kaggle/input/** — read-only mount for attached datasets (does not count against your 20 GB disk)
- **/kaggle/working/** — 20 GB writable workspace; output files saved here persist after commit
- **Kaggle Datasets** — access 50,000+ community datasets; attach any dataset to your notebook with one click via "Add Data"
- **Kaggle Models** — 200+ pre-trained model weights available for direct import
- **Private datasets** — upload your own data (up to 100 GB per dataset) and attach to notebooks
- **Tip:** compress large outputs with zip/tar to stay within the 500-file and 20 GB limits

---

## Pre-Installed Environment

Kaggle notebooks come with a comprehensive pre-configured environment:

- **Python 3.10+** with NumPy, Pandas, Scikit-learn, Matplotlib, Seaborn
- **Deep Learning:** TensorFlow, PyTorch, Keras, JAX, XGBoost, LightGBM, CatBoost
- **NLP:** HuggingFace Transformers, Tokenizers, Datasets
- **Computer Vision:** OpenCV, Albumentations, torchvision
- **R notebooks** also supported
- Install additional packages with `!pip install` (internet must be enabled)

---

## Practical Use Cases

- **Fine-tuning LLMs** — 16–32 GB VRAM is enough for QLoRA fine-tuning of 7B–13B parameter models
- **Training image classifiers** — P100 or T4 handles standard ResNet/EfficientNet training comfortably
- **Kaggle competitions** — the platform is purpose-built for this; many competition winners train exclusively on Kaggle GPUs
- **Prototyping and experimentation** — zero setup, just open a notebook and start coding
- **Running inference on large models** — load HuggingFace models directly from Kaggle Models

---

## Limitations

- **No persistent VM** — each session starts fresh; only `/kaggle/working/` outputs survive via commits
- **No SSH or terminal access** — you work through the notebook interface only
- **No deployment tools** — Kaggle is for experimentation, not production serving
- **Queue wait times** — during peak demand, GPU/TPU allocation may be delayed
- **Older GPUs** — P100 (2016) and T4 (2018) are capable but not cutting-edge; no A100 or H100 access
- **Session time limits** — 9-hour GPU sessions may not be enough for very large training runs
- **Floating quota** — your actual weekly hours may be less than 30 during high-demand periods

---

## Tips for Maximizing Free GPU Time

- **Use background execution (commit)** for all training runs — don't rely on keeping a browser tab open
- **Enable mixed precision (AMP)** when using T4 GPUs to leverage Tensor Cores
- **Save checkpoints frequently** to `/kaggle/working/` so you can resume across sessions
- **Chain notebooks** — save model checkpoints as a Kaggle dataset, then load them in a new notebook to continue training
- **Use the Kaggle API** (`kaggle kernels push`) to automate notebook execution from your local machine or CI/CD
- **Combine with Google Colab** — use KaggleHub to share datasets between Kaggle and Colab for additional free GPU hours
- **Monitor your quota** — check remaining GPU/TPU hours in the notebook Settings panel

---

## GitHub Integration

Kaggle allows you to connect your GitHub repository to upload and sync notebooks, making it easy to version-control your work outside the platform.

---

**Sources:**
- [Kaggle Notebooks](https://www.kaggle.com/code)
- [Kaggle Notebooks Documentation](https://www.kaggle.com/docs/notebooks)
- [Kaggle TPU Documentation](https://www.kaggle.com/docs/tpu)
- [How to Use GPU in Kaggle — GeeksforGeeks](https://www.geeksforgeeks.org/machine-learning/how-to-use-gpu-in-kaggle/)
- [Kaggle GPU Resources Guide — Oreate AI](https://www.oreateai.com/blog/a-comprehensive-guide-to-using-free-gpu-resources-on-kaggle/2a5a7cb02533e791828a210bdae44933)
- [Kaggle 29GB RAM GPU Update](https://medium.com/@fareedkhandev/kaggles-new-29gb-ram-gpus-the-power-you-need-absolutely-free-b458c3c501ba)
- [Gradient vs Kaggle Comparison — Paperspace](https://blog.paperspace.com/gradient-kaggle-notebook-comparison/)
- [Free Cloud GPU Comparison 2026 — AI Multiple](https://research.aimultiple.com/free-cloud-gpu/)
- [Free GPU Cloud Trials 2026 — GMI Cloud](https://www.gmicloud.ai/blog/where-can-i-get-free-gpu-cloud-trials-in-2025-a-complete-guide)
- [T4 vs P100 When to Choose — Medium](https://medium.com/@anagha.srivasa/nvidia-t4-x2-v-s-p100-gpu-when-to-choose-which-one-87cf1c55f386)

## Submission 42: JetBrains Student Pack + AI Free Tier

- Category: Education & Student
- Source Link: https://www.jetbrains.com/academy/student-pack
- Amount: Full JetBrains All Products Pack (normally ~$289/year for individuals) plus AI Free tier with unlimited local code completion and limited cloud AI credits
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Students and teachers at accredited educational institutions get **free access to all JetBrains professional IDEs** — IntelliJ IDEA Ultimate, PyCharm Professional, WebStorm, CLion, Rider, PhpStorm, DataGrip, RubyMine, GoLand, RustRover, and DataSpell — plus .NET tools like ReSharper. The license is valid for one year and renewable annually as long as you remain enrolled. Starting with IDE version 2025.1, all licensed users (including educational license holders) also get the **AI Free tier**: unlimited local code completion powered by JetBrains' Mellum model, offline AI via Ollama or LM Studio, and a small monthly cloud credit quota for AI chat and code generation using models like GPT-5.2, Claude 4.5 Sonnet, and Gemini. No credit card required.

---

## Registration (Step-by-Step)

1. Go to [jetbrains.com/community/education/#students](https://www.jetbrains.com/community/education/#students)
2. Choose your verification method (see below) and fill out the application form
3. Check your inbox and click the **confirmation link** in the verification email
4. Once approved, log in to your **JetBrains Account** at [account.jetbrains.com](https://account.jetbrains.com)
5. Download **JetBrains Toolbox App** (recommended) or individual IDEs directly
6. Activate your license by signing into the IDE with your JetBrains Account
7. AI Free tier activates automatically in IDEs version 2025.1+ — go to **Settings > AI Assistant** to configure

**Important:**
- No credit card is required at any point
- The license is **non-commercial only** — you cannot use it for paid work, freelancing, or commercial projects
- First-time AI activation includes a **free 30-day trial of AI Pro** with extended cloud quota; after that, you revert to AI Free
- After graduation, you get **40% off** all JetBrains products for the next two years

---

## Verification Methods

JetBrains offers three ways to prove student or teacher status:

| Method | Speed | Notes |
|---|---|---|
| **University email** (.edu, .ac.uk, etc.) | Instant / minutes | Fastest option. JetBrains maintains a list of recognized educational domains via the [SWOT repository](https://github.com/JetBrains/swot). If your domain is recognized, approval is automatic |
| **ISIC / ITIC card** | Hours to days | ISIC (students) or ITIC (teachers). Enter your card number in the application form. Get an ISIC card at [isic.org](https://www.isic.org/) if you do not have one |
| **GitHub Student Developer Pack** | Instant | If you already have the [GitHub Student Developer Pack](https://education.github.com/pack), select this option for automatic approval |

**Note:** As of July 2024, the "official document upload" option has been **discontinued**. If your university email domain is not recognized and you do not have ISIC or GitHub Student Pack, consider requesting your domain be added to the [SWOT repository](https://github.com/JetBrains/swot) via a pull request.

---

## Included IDEs and Tools

The Student Pack includes the full **All Products Pack** — every JetBrains desktop IDE and tool:

| IDE | Primary Language / Purpose |
|---|---|
| **IntelliJ IDEA Ultimate** | Java, Kotlin, Scala, Spring |
| **PyCharm Professional** | Python, Django, Flask, data science |
| **WebStorm** | JavaScript, TypeScript, React, Angular, Vue |
| **CLion** | C, C++, CMake, embedded |
| **Rider** | .NET, C#, Unity, Unreal Engine |
| **PhpStorm** | PHP, Laravel, Symfony |
| **GoLand** | Go |
| **RubyMine** | Ruby, Rails |
| **RustRover** | Rust |
| **DataGrip** | SQL, databases |
| **DataSpell** | Data science, Jupyter notebooks |
| **Aqua** | Test automation |
| **ReSharper** | .NET productivity (Visual Studio extension) |
| **ReSharper C++** | C++ productivity (Visual Studio extension) |
| **dotCover, dotMemory, dotTrace, dotPeek** | .NET profiling and analysis |

---

## AI Free Tier — What You Get

### Unlimited (no credits consumed)

- **Local code completion** — powered by JetBrains' **Mellum** model, a proprietary LLM specifically tuned for code completion. Works out of the box, no setup required
- **Next edit suggestions** — AI predicts your next likely edit location and change
- **Local AI chat** — connect your own models via **Ollama**, **LM Studio**, **llama.cpp**, **LiteLLM**, or any OpenAI-compatible API endpoint. Fully offline, no quota consumed
- **Local model code generation** — use models like Qwen2.5-Coder, DeepSeek-Coder, Codestral, or the open-source Mellum model locally

### Cloud credits (limited monthly quota)

- **AI chat** with cloud models (GPT-5.2, GPT-4.1, Claude 4.5 Sonnet, Claude 4.5 Haiku, Gemini 2.5 Pro/Flash, and more)
- **Smart in-editor suggestions** powered by cloud models
- **Junie** coding agent (limited usage)
- **Commit message generation**, documentation suggestions, test generation

The exact number of free-tier cloud credits is not publicly disclosed by JetBrains but is described as a "small" monthly quota. 1 AI Credit = $1.00 USD equivalent. The quota resets every 30 days from activation. You can check your remaining credits in the AI widget inside the IDE.

### Not available on AI Free

- Extended cloud quotas (requires AI Pro at $10/month or AI Ultimate at $20/month)
- Unlimited Junie agent usage

---

## Available AI Models

### Via JetBrains AI Service (cloud, consumes credits)

| Provider | Models |
|---|---|
| **OpenAI** | GPT-5.2, GPT-5, GPT-5 mini/nano, GPT-4.1, GPT-4o, o1, o3, o4-mini |
| **Anthropic** | Claude 4.5 Sonnet, Claude 4.5 Haiku, Claude 3.7 Sonnet |
| **Google** | Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash Lite, Gemini 2.0 Flash |
| **JetBrains** | Mellum (code completion, does not consume credits) |

### Via Local Models (unlimited, no credits)

| Host | Supported Models (examples) |
|---|---|
| **Ollama** | Qwen2.5-Coder, DeepSeek-Coder, CodeLlama, Mistral, Llama 3, Mellum |
| **LM Studio** | Any GGUF model |
| **llama.cpp / LiteLLM** | Any OpenAI-compatible endpoint |

Configure local models in **Settings > AI Assistant > Models**. Note: Junie agent and next edit suggestions do not support local models.

---

## IDEs Free for Non-Commercial Use (No Student Status Needed)

Several JetBrains IDEs are now **free for everyone** for non-commercial use, regardless of student status:

| IDE | Free Since |
|---|---|
| **RustRover** | Launch (2024) |
| **Aqua** | Launch (2024) |
| **WebStorm** | October 2024 |
| **Rider** | October 2024 |
| **CLion** | 2025 |
| **RubyMine** | 2025 |
| **DataGrip** | October 2025 |

These non-commercial licenses include full IDE features and AI Free tier access. The Student Pack is still valuable because it covers **IntelliJ IDEA Ultimate**, **PyCharm Professional**, **PhpStorm**, **GoLand**, and **DataSpell**, which are not free for non-commercial use.

---

## License Restrictions and Gotchas

- **Non-commercial only** — educational licenses cannot be used for commercial projects, freelancing, or paid work. If you need to do commercial work, you must purchase a personal or business license
- **Annual renewal required** — you must re-verify your student/teacher status every year. JetBrains sends a reminder email before expiration
- **One license per person** — the license is non-transferable
- **IDE version 2025.1+ required** for AI Free tier — older IDE versions do not include AI features
- **AI Free is NOT available in Community Editions** — PyCharm Community and IntelliJ IDEA Community do not support AI Free. You need the Professional/Ultimate editions (which the Student Pack provides)
- **Cloud AI credits are small** — if you rely heavily on cloud AI chat, you will likely exhaust the free quota quickly. Consider using local models via Ollama for unlimited AI usage

---

## Additional Tips

- **GitHub Student Developer Pack** is the easiest path if your university email is not recognized — apply at [education.github.com/pack](https://education.github.com/pack) first, then use it to get JetBrains access
- **Maximize AI usage with local models** — install [Ollama](https://ollama.com/) and download a coding model like `qwen2.5-coder:7b` for unlimited, private AI assistance without spending cloud credits
- **JetBrains Toolbox App** — install this first to manage all your IDE installations, updates, and licenses from one place
- **JetBrains Academy (Hyperskill)** — students also get access to interactive programming courses through JetBrains Academy as part of the Student Pack
- **Open source license** — if you maintain an active open source project, you can apply for a separate [free open source license](https://www.jetbrains.com/community/opensource/) that allows broader use
- **Startup program** — if you are building a startup, JetBrains offers a [startup program](https://www.jetbrains.com/community/startup/) with free licenses
- **30-day AI Pro trial** — when you first enable AI, you automatically get 30 days of AI Pro. Use this trial to explore cloud-powered features before they revert to the smaller free quota

---

**Sources:**
- [JetBrains Student Pack](https://www.jetbrains.com/academy/student-pack/)
- [JetBrains Educational License FAQ](https://sales.jetbrains.com/hc/en-gb/articles/207241195-Do-you-offer-free-educational-licenses-for-students-and-teachers)
- [How to Apply for Educational License](https://sales.jetbrains.com/hc/en-gb/articles/11558649766674-How-do-I-apply-for-a-free-educational-license)
- [JetBrains AI Assistant 2025.1 Release](https://blog.jetbrains.com/ai/2025/04/jetbrains-ai-assistant-2025-1-2/)
- [JetBrains AI Licensing and Subscriptions](https://www.jetbrains.com/help/ai-assistant/licensing-and-subscriptions.html)
- [JetBrains AI FAQ](https://lp.jetbrains.com/ai-ides-faq/)
- [AI Quota Model Update (August 2025)](https://blog.jetbrains.com/ai/2025/08/a-simpler-more-transparent-model-for-ai-quotas/)
- [JetBrains AI Supported LLMs](https://www.jetbrains.com/help/ai-assistant/supported-llms.html)
- [AI for Students — Knowledge Base](https://youtrack.jetbrains.com/articles/SUPPORT-A-395/Is-AI-Assistant-included-available-in-the-IDE-students-educational-pack-subscription)
- [JetBrains Student Pack Blog Post](https://blog.jetbrains.com/education/2025/08/12/jetbrains-student-pack/)
- [Ollama JetBrains Integration](https://docs.ollama.com/integrations/jetbrains)
- [Non-Commercial Licensing FAQ](https://sales.jetbrains.com/hc/en-gb/articles/18950890312210-The-free-non-commercial-licensing-FAQ)
- [JetBrains SWOT Repository](https://github.com/JetBrains/swot)

## Submission 41: IBM watsonx.ai Lite Plan (Free Tier)

- Category: Cloud Providers
- Source Link: https://www.ibm.com/products/watsonx-ai/pricing
- Amount: 20 CUH/month + 300,000 tokens/month
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Always-free tier for IBM's enterprise AI platform providing **20 Capacity Unit Hours (CUH) per month** for ML workloads and **300,000 tokens per month** for foundation model inference. No credit card charge -- identity verification only. Includes access to IBM Granite models, Meta Llama, Mistral, and other third-party foundation models. Best suited for experimentation, prototyping, and learning.

---

## Registration (Step-by-Step)

1. Go to [dataplatform.cloud.ibm.com/registration/stepone?context=wx](https://dataplatform.cloud.ibm.com/registration/stepone?context=wx)
2. Select an **IBM Cloud region** (Dallas or Frankfurt recommended for full foundation model access)
3. Log in with your **IBMid** or create a new IBM Cloud account (email, personal info, and credit card for identity verification)
4. If prompted, select your **account and resource group**
5. Click **Continue** and wait for activation to complete
6. Bookmark your watsonx.ai home page for future logins

**Notes:**
- One Lite plan instance per IBM Cloud account
- If you land on the IBM Cloud Dashboard instead, return to the signup page and click "Log in with existing account"
- No charges unless you manually upgrade to a paid plan

---

## What Is a Capacity Unit Hour (CUH)?

CUH is IBM's billing unit for all watsonx.ai Runtime activity **except** foundation model inference. CUH is consumed when you:

- Train **AutoAI** models
- Run or score **deployed ML models**
- Execute **Decision Optimization** jobs
- Tune models in the **Tuning Studio** (not available on Lite)

CUH consumption depends on the training tool, hardware spec, and runtime environment. Compute time is measured to the millisecond with a **one-minute minimum** per operation. A deployment that receives no scoring requests for a set duration goes idle and stops consuming CUH.

**Example:** A Decision Optimization batch job running 15 minutes on 2 nodes (2 vCPU, 8 GB RAM, CUH rate 30) consumes 0.25 x 2 x 30 = **15 CUH**.

Foundation model inference uses a separate metric: **Resource Units (RU)**, where 1 RU = 1,000 tokens.

---

## Lite Plan Limits

| Resource | Lite Plan Limit |
|---|---|
| **CUH** | 20 per month |
| **Foundation model tokens** | 300,000 per month |
| **Inference request rate** | 2 requests/second |
| **Deployment idle timeout** | 1 day |
| **Parallel batch jobs (Decision Opt.)** | 2 per deployment |
| **Deployment job retention** | 100 jobs per space |
| **Document text classification** | 100 pages/month |
| **Cloud Object Storage** | 25 GB/month |
| **Studio instances** | 1 per account |

---

## Available Foundation Models

The Lite plan provides access to the same foundation models as paid plans, with capacity limits. Model availability varies by data center.

### IBM Granite Models

| Model | Parameters | Use Cases |
|---|---|---|
| **granite-4-h-small** | Small | General purpose |
| **granite-3-3-8b-instruct** | 8B | Instruction-following, Q&A, summarization |
| **granite-3-8b-base** | 8B | Base model for fine-tuning |
| **granite-8b-code-instruct** | 8B | Code generation and explanation |
| **granite-guardian-3-8b** | 8B | Safety filtering, HAP/PII detection |
| **granite-ttm variants** | Various | Time-series forecasting |

### Meta Llama Models

| Model | Parameters | Use Cases |
|---|---|---|
| **llama-4-maverick-17b-128e-instruct-fp8** | 17B (128 experts) | Multimodal, latest generation |
| **llama-3-3-70b-instruct** | 70B | Large-scale instruction-following |
| **llama-3-2-11b-vision-instruct** | 11B | Multimodal vision + text |
| **llama-guard-3-11b-vision** | 11B | Safety classification |

### Mistral AI Models

| Model | Notes |
|---|---|
| **mistral-large-2512** | 256K context window |
| **mistral-medium-2505** | Medium capability tier |
| **mistral-small-3-1-24b-instruct** | 24B, efficient small model |

### Other Models

| Model | Provider | Notes |
|---|---|---|
| **gpt-oss-120b** | OpenAI | Open-source variant, 120B parameters |
| **mt0-xxl-13b** | BigScience | Multilingual translation |
| **allam-1-13b-instruct** | SDAIA | Arabic language model |

---

## What Can You Build with 300K Tokens/Month?

A token is roughly **4 characters or 0.75 words**. With 300,000 tokens per month (input + output combined):

| Use Case | Approx. Capacity |
|---|---|
| Short Q&A prompts (~200 tokens each) | ~1,500 interactions |
| Summarizing documents (~1,000 tokens each) | ~300 documents |
| Code generation (~500 tokens each) | ~600 code snippets |
| RAG-based chatbot (~800 tokens per exchange) | ~375 conversations |
| Small classification pipeline (~100 tokens each) | ~3,000 classifications |

This is enough for prototyping, academic projects, demos, and light personal use. Not suitable for production workloads.

---

## Lite Plan Restrictions

The Lite plan does **not** support:

- **Foundation model tuning** (prompt tuning, fine-tuning)
- **Custom foundation model deployments**
- **GPU runtime environments** for notebooks
- **Large runtime environments** (8+ vCPU)
- **Project export**
- **Encryption with your own key** (BYOK)
- **Quota increases** beyond Lite limits

Collaborators in your projects must each have their own Lite plan.

---

## Tips for Maximizing the Free Tier

- **Use smaller models** (granite-4-h-small, mistral-small) to consume fewer tokens per interaction
- **Monitor usage** in the IBM Cloud dashboard -- track CUH and RU consumption to avoid hitting limits mid-month
- **Use the Prompt Lab** for interactive testing before building pipelines
- **Select Dallas or Frankfurt** as your region for the widest model availability
- **Idle deployments** time out after 1 day -- redeploy as needed rather than keeping deployments running
- **Combine with the free demo** at [ibm.com/products/watsonx/get-started](https://www.ibm.com/products/watsonx/get-started) for an additional 20,000 tokens over 30 days (requires only IBM/Google/LinkedIn login)
- **Use the Python SDK** (`ibm-watsonx-ai`) for programmatic access and automation
- **Set up billing alerts** if you ever upgrade, as there are no spending caps by default

---

## Upgrading Options

| Plan | Price | Notes |
|---|---|---|
| **Lite** | Free | 20 CUH + 300K tokens/month |
| **Essentials** | Pay-as-you-go (from $0/month) | No standing charge, pay only for usage |
| **Standard** | ~$1,050/month base | Enterprise support, higher limits |
| **Professional** | Custom pricing | Flat-rate, full capacity |

---

**Sources:**
- [IBM watsonx.ai Pricing](https://www.ibm.com/products/watsonx-ai/pricing)
- [watsonx.ai Runtime Service Plans](https://dataplatform.cloud.ibm.com/docs/content/wsj/getting-started/wml-plans.html?context=wx)
- [Supported Foundation Models](https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx)
- [Signing Up for IBM watsonx](https://dataplatform.cloud.ibm.com/docs/content/wsj/getting-started/signup-wx.html?context=wx)
- [Billing Details for ML Assets](https://dataplatform.cloud.ibm.com/docs/content/wsj/getting-started/wxai-runtime-plans-ml.html?context=wx)
- [Billing Details for GenAI Assets](https://dataplatform.cloud.ibm.com/docs/content/wsj/getting-started/wxai-runtime-plans-genai.html?context=wx)
- [IBM watsonx.ai Foundation Models](https://www.ibm.com/products/watsonx-ai/foundation-models)
- [Try IBM watsonx.ai for Free](https://dataplatform.cloud.ibm.com/registration/stepone?context=wx)

## Submission 40: IBM Academic Initiative & SkillsBuild

- Category: Education & Student
- Source Link: https://www.ibm.com/academic
- Amount: Free IBM software suite, IBM Cloud trial (6-12 months), 1,000+ courses, and digital credentials
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

IBM offers **two interconnected free programs** for students and educators at accredited institutions: **IBM SkillsBuild Technology Access** (formerly IBM Academic Initiative) provides free downloads of full-version IBM software, IBM Cloud trial accounts with feature codes, and development tools. **IBM SkillsBuild** adds 1,000+ free online courses, hands-on AI labs, and shareable digital credentials via Credly. No credit card required. Best for students who want enterprise-grade IBM tools for coursework, research, or personal projects.

---

## Registration (Step-by-Step)

### IBM SkillsBuild Technology Access (Software & Cloud)

1. Go to [ibm.com/academic](https://www.ibm.com/academic/) and click **"Access software downloads"**
2. Enter your **institutional email address** (e.g., `you@university.edu`)
3. Select your role: **Student** or **Faculty**
4. Complete the registration form with your personal and institutional details
5. Optionally enroll in the **IBM Badge Program** for digital credentials
6. Review and accept the **IBM SkillsBuild Technology Access Agreement** and Privacy Statement
7. Click **Register** to submit
8. Sign in with an existing **IBMid** linked to your institutional email, or create a new IBMid
9. Return to the portal and log in — you now have access to software downloads and can request IBM Cloud feature codes

### IBM SkillsBuild (Courses & Credentials)

1. Go to [skillsbuild.org](https://skillsbuild.org/) and click **"Get started"**
2. Choose **"College students"** or **"Learners"** depending on your status
3. Create an account with your email (institutional email not strictly required for courses, but recommended)
4. Browse the course catalog and enroll in any course for free

**Important:**
- One Academic Initiative account per person
- You must use an email issued by a **degree-granting accredited institution**
- Registration takes approximately 5-10 minutes
- Software access is for **classroom, learning, and non-commercial research** purposes only

---

## IBM Cloud Feature Code

The most valuable part for builders is the **IBM Cloud Feature Code**, which upgrades your IBM Cloud Lite account to a Trial account with expanded access.

### How to Request a Feature Code

1. Log in to [academic.ibm.com](https://academic.ibm.com/a2mt/email-auth)
2. Navigate to **Cloud > Software > IBM Cloud Feature Code**
3. Click **"Request Feature Code"**
4. Copy the 32-character alphanumeric code (also emailed to you)
5. Log in to [cloud.ibm.com](https://cloud.ibm.com)
6. Go to **Manage > Account > Account settings > Subscription and Feature codes**
7. Apply the code

### What the Feature Code Provides

| Feature | Details |
|---|---|
| **Account type** | Upgraded from Lite to Trial |
| **Duration** | Up to 6 months for students (ending June 30 or Dec 31); 12 months for faculty |
| **Memory** | Increased allocation up to 2 GB |
| **Services** | Access to free tier ($Free Plan) of all IBM Cloud services |
| **Credit card** | Not required |
| **Renewable** | Yes — request a new code before expiration |

**Gotchas:**
- If you add a credit card, your account upgrades to Pay-As-You-Go and **cannot be converted back** to Trial
- When the trial expires without renewal, the account is **deactivated and resources deleted**
- The same feature code cannot be applied twice — request a new one each period

---

## Available Software Downloads

IBM provides **full-version software** (not limited academic editions) through the software catalog at [academic.ibm.com/a2mt/downloads](https://academic.ibm.com/a2mt/downloads). The catalog is organized into technology areas including:

| Category | Examples |
|---|---|
| **Cloud** | IBM Cloud Pak offerings, Cloud Foundry tools |
| **Data & Analytics** | IBM SPSS Statistics, SPSS Modeler, Cognos Analytics |
| **AI / Watson** | Watson Studio, Watson Machine Learning tools |
| **Security** | IBM QRadar, Guardium tools |
| **Automation** | IBM Robotic Process Automation |
| **Quantum** | IBM Quantum (Qiskit) resources |
| **Mainframe / IBM Z** | z/OS development tools, COBOL compilers |
| **Integration** | IBM MQ, App Connect |

**Note:** The full catalog is only visible after logging in with a verified academic email. Software is licensed for educational and non-commercial research use only.

---

## IBM Cloud Free Tier Services (with Academic Account)

With your Trial account, you get access to the same free-tier services available to all IBM Cloud users, plus expanded access through the feature code:

| Service | Free Limit |
|---|---|
| **Watson Assistant** | 10,000 API calls/month |
| **Watson Discovery** | 1,000 documents/month |
| **Speech to Text** | 500 minutes/month |
| **Text to Speech** | 10,000 characters/month |
| **Language Translator** | 1,000,000 characters/month |
| **Natural Language Understanding** | 30,000 API calls/month |
| **Cloud Object Storage** | 25 GB/month |
| **watsonx.ai Lite** | 300,000 tokens/month + 20 CUH/month |
| **Db2** | 200 MB storage |
| **Cloudant (NoSQL)** | 1 GB storage |

---

## IBM SkillsBuild Courses & Labs

Over 1,000 free courses are available, with no institutional email required for the learning platform itself. Key AI and tech courses include:

### AI & Machine Learning

| Course | Duration | Credential |
|---|---|---|
| Getting Started with Artificial Intelligence | 3 hours | Badge |
| Getting Started with Generative AI | 3 hours | Badge |
| AI Literacy | 4 hours | Badge |
| Artificial Intelligence Fundamentals | 10+ hours | Badge |
| Prompt Engineering: Shaping Better AI Responses | 4 hours | — |
| Retrieval-Augmented Generation for Enhanced AI Outputs | 4 hours | — |
| Make Agentic AI Work for You | 4 hours | — |
| Introduction to Large Language Models | 90 min | — |
| Ethical Considerations for Using Generative AI | 60-90 min | — |

### Hands-On AI Labs

- Prompt a Foundation Model in IBM watsonx
- Document Summarization with IBM Granite
- Build AI-Powered Multimodal RAG Systems
- Implement RAG Using LlamaIndex
- Build Your First Chatbot Using IBM watsonx

### Other Technical Tracks

| Track | Duration | Credential |
|---|---|---|
| Cloud Computing Fundamentals | 10 hours | Badge |
| Cybersecurity Fundamentals | 7.5 hours | Badge |
| Data Fundamentals | 7 hours | Badge |
| Web Development Fundamentals | 12+ hours | Badge |
| Information Technology Fundamentals | 11 hours | Badge |
| Project Management Fundamentals | 7 hours | Badge |
| Quantum Enigmas | 8 hours | Badge |
| Enterprise Design Thinking Practitioner | 3 hours | Badge |

---

## Digital Credentials (Badges)

Completing SkillsBuild courses earns you **digital badges** issued through [Credly](https://www.credly.com/org/ibm-skillsbuild). Each badge contains metadata about your verified skills and can be shared on LinkedIn, resumes, and portfolios.

To earn a badge, you must complete the prescribed coursework and **pass all assessments with 80% or higher**.

Key AI-related badges:
- **Artificial Intelligence Fundamentals** — NLP, computer vision, ML, deep learning, neural networks
- **Getting Started with Artificial Intelligence** — Generative AI tools and prompt creation
- **AI Literacy** — How AI works, ethics, and applications
- **Cloud Computing Fundamentals** — Cloud services, deployment models, security
- **Data Fundamentals** — Data cleaning, refinement, and visualization with Watson Studio
- **Explore Emerging Tech** — Data, AI, cloud, cybersecurity, and quantum computing

---

## Access Duration & Renewal

| Component | Duration | Renewable |
|---|---|---|
| **Software Downloads** | 6 months (students) / 12 months (faculty) | Yes, while university email is valid |
| **IBM Cloud Feature Code** | Up to 6 months per code | Yes, request new code before expiration |
| **SkillsBuild Courses** | Ongoing, no expiration | N/A |
| **Digital Badges** | Permanent once earned | N/A |

---

## Virtual Internship Program

IBM SkillsBuild also offers a **free Virtual Internship Program** where students complete guided coursework and hands-on tasks in tracks like AI, ML, Cloud Computing, Cybersecurity, and Data Analytics. Successful participants receive a digital certificate and, where applicable, a digital badge.

---

## Additional Tips

- **SkillsBuild courses are open to everyone** — you do not need a .edu email. But software downloads and IBM Cloud feature codes require verified institutional email
- **Do not add a credit card** to your IBM Cloud trial account — it permanently upgrades to Pay-As-You-Go and you lose trial benefits
- **Request a new feature code before your current one expires** to avoid account deactivation and resource deletion
- **IBM Z Student Ambassador Program** — a separate application-based program for students interested in mainframe computing, with mentorship and networking
- **HBCU partnerships** — IBM provides micro-internships and mentors through SkillsBuild at participating HBCUs
- **Combine with watsonx.ai Lite** — the always-free watsonx.ai Lite plan (300K tokens/month) works independently of the Academic Initiative and does not require an .edu email
- **Use the hands-on labs** — the watsonx AI labs are the fastest way to get practical experience with foundation models like IBM Granite
- **Badges are employer-recognized** — IBM digital credentials on Credly are used in hiring processes and can differentiate your profile

---

**Sources:**
- [IBM Academic Initiative / SkillsBuild Portal](https://www.ibm.com/academic/)
- [IBM SkillsBuild for College Students](https://skillsbuild.org/college-students)
- [IBM SkillsBuild Course Catalog](https://skillsbuild.org/college-students/course-catalog)
- [IBM SkillsBuild Digital Credentials](https://skillsbuild.org/students/digital-credentials)
- [IBM SkillsBuild Software Downloads](https://academic.ibm.com/a2mt/downloads)
- [Registration Guide (GitHub)](https://github.com/academic-initiative/documentation/blob/main/academic-initiative/how-to/How-to-register-with-the-IBM-Academic-Initiative/readme.md)
- [IBM Cloud Feature Code Guide (GitHub)](https://github.com/academic-initiative/documentation/blob/main/academic-initiative/how-to/How-to-request-and-IBM-Cloud-Feature-Code/readme.md)
- [IBM Cloud Feature Code FAQ](https://www.ibm.com/support/pages/how-do-i-get-feature-code-sign-academic-initiative)
- [IBM SkillsBuild on Credly](https://www.credly.com/org/ibm-skillsbuild)
- [IBM Academic Initiative Overview](https://www.ibm.com/ibm/responsibility/academic_initiative.shtml)

## Submission 39: Hugging Face - Free Inference API Credits

- Category: AI API Free Tiers
- Source Link: https://huggingface.co/inference-api
- Amount: ~$0.10/month in free inference credits (subject to change)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Every Hugging Face user receives **$0.10/month in free inference credits** (subject to change) to experiment with Inference Providers — a unified API that routes requests to 200+ models across 18+ inference partners. No credit card required. Free users cannot continue past their monthly credit limit (no pay-as-you-go). The PRO plan ($9/month) bumps credits to **$2/month** and unlocks pay-as-you-go billing after credits run out. Hugging Face charges provider rates with **zero markup**.

---

## Registration (Step-by-Step)

1. Go to [huggingface.co/join](https://huggingface.co/join)
2. Sign up with **email** or sign in with **Google**, **GitHub**, or other OAuth providers
3. Confirm your email address
4. Go to [Token Settings](https://huggingface.co/settings/tokens/new?ownUserPermissions=inference.serverless.write&tokenType=fineGrained) and create a **fine-grained token** with the **"Make calls to Inference Providers"** permission
5. Done — you can now use the Inference API immediately with your $0.10/month free credits

**Important:**
- No credit card required at any stage
- No approval process — access is instant
- Free credits reset monthly; unused credits do not roll over
- When credits run out, you get a `402 Payment Required` error — free users cannot continue past the limit

---

## How Inference Providers Work

Hugging Face Inference Providers is a **proxy layer** that sits between your app and multiple AI providers (Groq, Together AI, SambaNova, Fireworks, Replicate, etc.). You send requests using a single Hugging Face token and the system routes them to the best available provider.

Two billing modes:

| Mode | How It Works | Credits Apply? | Pay-as-you-go? |
|---|---|---|---|
| **Routed by Hugging Face** (default) | Request routes through HF to a provider | Yes | Only for PRO/Enterprise |
| **Custom Provider Key** | You use your own provider API key | No | Billed by provider directly |

The API is **OpenAI-compatible** — you can swap in `https://router.huggingface.co/v1` as a base URL with any OpenAI SDK.

---

## Available Models & Providers

Access 200+ models across these providers:

| Provider | LLMs | Vision LLMs | Embeddings | Text-to-Image | Text-to-Video | Speech-to-Text |
|---|---|---|---|---|---|---|
| **Cerebras** | Yes | | | | | |
| **Cohere** | Yes | Yes | | | | |
| **Fal AI** | | | | Yes | Yes | Yes |
| **Fireworks** | Yes | Yes | | | | |
| **Groq** | Yes | Yes | | | | |
| **HF Inference** | Yes | Yes | Yes | Yes | | Yes |
| **Hyperbolic** | Yes | Yes | | | | |
| **Novita** | Yes | Yes | | | Yes | |
| **Nscale** | Yes | Yes | | Yes | | |
| **Replicate** | | | | Yes | Yes | Yes |
| **SambaNova** | Yes | | Yes | | | |
| **Together AI** | Yes | Yes | | Yes | | |

Popular models include **DeepSeek-R1**, **DeepSeek-V3**, **Llama 3/4 family**, **Mistral/Mixtral**, **Qwen 2.5/3**, **FLUX.1 (image gen)**, **GPT-OSS-120B**, and many more. Browse the full list at [huggingface.co/inference/models](https://huggingface.co/inference/models).

---

## Provider Selection & Routing

You can control which provider handles your request:

- **Automatic** (default): Routes to the first available provider based on your preference order
- **`:fastest` suffix**: Selects the provider with highest throughput (e.g., `deepseek-ai/DeepSeek-R1:fastest`)
- **`:cheapest` suffix**: Selects the provider with lowest price per output token
- **Explicit provider**: Specify directly (e.g., `deepseek-ai/DeepSeek-R1:sambanova`)

---

## Free Credits vs. PRO Plan

| Feature | Free ($0) | PRO ($9/month) |
|---|---|---|
| **Monthly inference credits** | ~$0.10 (subject to change) | $2.00 |
| **Pay-as-you-go after credits** | No — hard stop at limit | Yes — billed at provider rates |
| **ZeroGPU Spaces usage** | ~300 seconds, low priority | 8x quota (~2,400 sec), highest priority |
| **ZeroGPU Spaces hosting** | Cannot host | Up to 10 Spaces (H200 GPU) |
| **Private storage** | 100 GB | 1 TB (10x) |
| **Queue priority** | Standard | Highest |

---

## ZeroGPU Spaces (Bonus Free GPU Access)

Separate from Inference Providers, Hugging Face offers **ZeroGPU Spaces** — public Gradio apps that dynamically allocate NVIDIA H200 GPUs on demand:

- **Free users** can use any public ZeroGPU Space with a rate limit of ~300 seconds per session (refills at 1 ZeroGPU second per 30 real seconds)
- **PRO users** get 8x the quota and highest queue priority
- Only PRO users and Enterprise orgs can **host** ZeroGPU Spaces; anyone can **use** them
- Currently only works with the Gradio SDK

---

## Billing & Pricing Details

- Hugging Face charges **zero markup** on provider rates — you pay exactly what the provider charges
- Billing is per-request based on **compute time x hardware cost per second**
- Example: A FLUX.1-dev image generation taking 10 seconds on a GPU at $0.00012/sec = $0.0012 per image
- Track spending at [huggingface.co/settings/billing](https://huggingface.co/settings/billing)
- Detailed per-model, per-provider breakdown at [Inference Providers Settings](https://huggingface.co/settings/inference-providers/overview)

---

## Integration (Quick Start)

The API works as a **drop-in OpenAI replacement**:

```python
from openai import OpenAI

client = OpenAI(
    base_url="https://router.huggingface.co/v1",
    api_key="hf_YOUR_TOKEN",
)

completion = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V3-0324",
    messages=[{"role": "user", "content": "Hello!"}],
)
```

Also available via the native `huggingface_hub` Python/JS SDK, direct HTTP/cURL, and the [Inference Playground](https://huggingface.co/playground) for browser-based testing.

---

## Additional Tips

- **$0.10 goes fast** — a few LLM chat completions or one image generation can exhaust your monthly free credits. Treat it as a trial, not a production budget
- **Limits have been tightened** — multiple community reports confirm that free tier limits were reduced in late 2024/early 2025. Users who previously ran hundreds of requests now hit limits much sooner
- **Bring your own key** — if you already have accounts with Groq (free tier), Together AI, or SambaNova, you can use their API keys through Hugging Face without consuming HF credits
- **Team plan** ($20/user/month) and **Enterprise** (from $50/user/month) provide $2/seat in shared credits plus centralized billing
- **No SLA on free tier** — there is no uptime or latency guarantee for free users
- **HF Inference provider** (the built-in one) focuses mostly on CPU inference as of mid-2025 — for GPU-accelerated LLMs, requests are routed to external providers like Groq, SambaNova, or Together AI
- **OpenAI-compatible endpoint** only supports chat completions; for image generation, embeddings, or speech tasks, use the native HF SDK

---

**Sources:**
- [Hugging Face Inference Providers Pricing](https://huggingface.co/docs/inference-providers/en/pricing)
- [Hugging Face Pricing Page](https://huggingface.co/pricing)
- [Hugging Face Inference Providers Documentation](https://huggingface.co/docs/inference-providers/en/index)
- [Hugging Face Supported Models](https://huggingface.co/inference/models)
- [Hugging Face PRO Plan](https://huggingface.co/subscribe/pro)
- [ZeroGPU Spaces Documentation](https://huggingface.co/docs/hub/en/spaces-zerogpu)
- [Inference Providers Blog Post](https://huggingface.co/blog/inference-providers)
- [Hugging Face Forum: Free Monthly Limit Reached](https://discuss.huggingface.co/t/free-monthly-limit-reached/171827)
- [Hugging Face Forum: API Inference Limit Changed](https://discuss.huggingface.co/t/api-inference-limit-changed/144157)

## Submission 38: Groq Free Tier - LPU-Powered Inference

- Category: AI API Free Tiers
- Source Link: https://console.groq.com/
- Amount: Rate-limited free usage (varies by model; ~30 RPM, ~6K-12K TPM, ~100K-500K TPD)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Groq offers a **permanent free tier** for its LPU (Language Processing Unit) powered inference API. Sign up at [console.groq.com](https://console.groq.com), get an API key instantly, and start calling models with **no credit card required**. The free tier gives you access to all supported models — including Llama 3.1/3.3, Llama 4 Scout/Maverick, GPT-OSS, Qwen3, Kimi K2, Whisper, and Groq's agentic Compound system — subject to rate limits (~30 RPM, ~6K TPM for smaller models). The API is **OpenAI-compatible**, so most libraries and tools that work with OpenAI also work with Groq by changing the base URL. Groq's main selling point is **speed**: inference runs on custom LPU hardware at 200-1,000+ tokens/second depending on the model.

---

## Registration (Step-by-Step)

1. Go to [console.groq.com](https://console.groq.com)
2. Click **"Sign Up"** (or use the direct link: [console.groq.com/authenticate/signup](https://console.groq.com/authenticate/signup))
3. Sign up with your **email address** or **GitHub account** — accept the Services Agreement and Privacy Policy
4. Verify your email if prompted
5. You are now on the **Free tier** — no credit card needed, no trial period
6. Go to the **API Keys** page and click **"Create API Key"**
7. Give your key a descriptive name and **copy it immediately** — it will not be shown again
8. Set it as an environment variable: `export GROQ_API_KEY=your-key-here`
9. Make your first request to `https://api.groq.com/openai/v1/chat/completions`

**Important:**
- One free account per person — no special verification required
- The free tier has **no expiration date** — it is ongoing, not a trial
- You will **never be charged** unless you explicitly add payment info and upgrade to the Developer tier
- Over 2M+ developers are already building on GroqCloud

---

## Available Models

### Production Models

| Model | Speed | Context Window | Pricing (paid tier) |
|---|---|---|---|
| **Llama 3.1 8B Instant** | ~840 t/s | 131,072 | $0.05/$0.08 per 1M tokens |
| **Llama 3.3 70B Versatile** | ~394 t/s | 131,072 | $0.59/$0.79 per 1M tokens |
| **GPT-OSS 120B** | ~500 t/s | 131,072 | $0.15/$0.60 per 1M tokens |
| **GPT-OSS 20B** | ~1,000 t/s | 131,072 | $0.075/$0.30 per 1M tokens |
| **Llama Guard 4 12B** | ~1,200 t/s | 131,072 | Content moderation model |
| **Whisper Large V3** | 217x realtime | — | $0.111/hour transcribed |
| **Whisper Large V3 Turbo** | 228x realtime | — | $0.04/hour transcribed |

### Preview Models (may be discontinued)

| Model | Notes |
|---|---|
| **Llama 4 Maverick 17Bx128E** | ~562 t/s, 128K context |
| **Llama 4 Scout 17Bx16E** | ~594 t/s, 128K context |
| **Qwen3 32B** | ~662 t/s, 131K context |
| **Kimi K2 (1T params)** | ~200 t/s, 256K context |
| **Llama Prompt Guard 2** | 22M and 86M variants |
| **PlayAI TTS** | Text-to-speech (English + Arabic) |

### Compound Systems (Agentic AI)

| System | Description |
|---|---|
| **groq/compound** | Agentic system with built-in web search + code execution; multiple tool calls per request |
| **groq/compound-mini** | Lighter version; single tool call per request, ~3x lower latency |

Compound systems can autonomously search the web and execute Python code server-side — no client-side orchestration needed. Set `model` to `groq/compound` in your API call and the system handles the rest.

**Note:** The model lineup changes frequently. Check the [official models page](https://console.groq.com/docs/models) for the current list.

---

## Free Tier Rate Limits

Rate limits apply at the **organization level** (not per user) and vary by model. Smaller models get higher limits. Here are representative free tier limits:

| Model | RPM | RPD | TPM | TPD |
|---|---|---|---|---|
| **Llama 3.1 8B Instant** | 30 | 14,400 | 6,000 | 500,000 |
| **Llama 3.3 70B Versatile** | 30 | 1,000 | 12,000 | 100,000 |
| **Allam 2 7B** | 30 | 7,000 | 6,000 | 500,000 |
| **Whisper Large V3** | 20 | 2,000 | — | — |
| **Whisper Large V3 Turbo** | 20 | 2,000 | — | — |

*RPM = Requests/Minute, RPD = Requests/Day, TPM = Tokens/Minute, TPD = Tokens/Day*

**Key details:**
- Limits reset daily at **UTC 00:00**
- **Cached tokens do not count** toward rate limits (prompt caching gives 50% discount on paid tier)
- Exceeding limits returns a **429 Too Many Requests** error — you are never charged, just throttled
- Max output per request: **8,192 tokens**
- The Developer tier offers roughly **10x higher limits** across all models
- Exact limits change frequently — check [your account limits](https://console.groq.com/settings/limits) for current numbers

---

## OpenAI-Compatible API

Groq's API is **drop-in compatible with the OpenAI API format**. This means:

- **Endpoint:** `https://api.groq.com/openai/v1/chat/completions`
- Works with the official **OpenAI Python/Node SDKs** — just change the `base_url` and API key
- Works with **LangChain**, **LiteLLM**, **AI SDK**, and most OpenAI-compatible tools
- Groq also provides its own Python SDK: `pip install groq`

```python
from groq import Groq
import os

client = Groq(api_key=os.environ.get("GROQ_API_KEY"))
chat = client.chat.completions.create(
    messages=[{"role": "user", "content": "Hello!"}],
    model="llama-3.3-70b-versatile",
)
print(chat.choices[0].message.content)
```

---

## Upgrading to Developer Tier

If the free tier limits are too restrictive, the **Developer tier** offers:

- Up to **10x higher rate limits** across all models
- Access to the **Batch API** for high-volume workloads (25% cost discount)
- Pay-as-you-go pricing — you only pay for tokens consumed
- Self-serve upgrade from the GroqCloud console

Pricing is very competitive: Llama 3.1 8B starts at **$0.05 per million input tokens**, and even the largest models (Kimi K2 at 1T params) cost only **$1.00 per million input tokens**.

---

## Additional Tips

- **Best for prototyping and experimentation** — the free tier is generous enough for hobby projects, hackathons, and testing, but not for production workloads
- **Speed is the main differentiator** — Groq's LPU hardware delivers inference speeds 5-10x faster than GPU-based providers. If latency matters, Groq is hard to beat
- **Implement retry logic** — on the free tier, you will hit rate limits. Use exponential backoff on 429 errors
- **Token counting** — both input and output tokens count toward TPM/TPD limits
- **No SLA on free tier** — models in Preview status may be removed without notice
- **Compound systems are billed per-tool-use** on paid tier — web search costs $5-8 per 1,000 requests, code execution $0.18/hour. On free tier, these are available within your rate limits
- **Prompt caching** — on the paid tier, cached tokens get a 50% discount. On free tier, cached tokens simply don't count toward your rate limits
- **Groq vs. GraphQL (GROQ)** — Groq (the AI inference company) is unrelated to GROQ (Sanity's query language). Don't confuse them

---

**Sources:**
- [GroqCloud Console](https://console.groq.com)
- [Groq Supported Models](https://console.groq.com/docs/models)
- [Groq Rate Limits Documentation](https://console.groq.com/docs/rate-limits)
- [Groq Pricing](https://groq.com/pricing)
- [Groq Quickstart Guide](https://console.groq.com/docs/quickstart)
- [Groq Community FAQ - Free Tier](https://community.groq.com/t/is-there-a-free-tier-and-what-are-its-limits/790)
- [Groq Community FAQ - Rate Limits](https://community.groq.com/t/what-are-the-rate-limits-for-the-groq-api-for-the-free-and-dev-tier-plans/42)
- [Groq Compound Systems](https://console.groq.com/docs/compound)
- [Groq Developer Tier Blog Post](https://groq.com/blog/developer-tier-now-available-on-groqcloud)

## Submission 37: Google for Startups Cloud Program - Start Tier ($2,000)

- Category: Startup Programs
- Source Link: https://cloud.google.com/startup
- Amount: $2,000
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Pre-funded startups can get **up to $2,000 in Google Cloud credits valid for 1 year** through the Google for Startups Cloud Program Start tier. This is designed for early-stage founders who have not yet raised equity funding and want to get started building on GCP and Firebase. No credit card hold or payment is taken for the credits themselves, but you do need a Google Cloud Billing Account (which requires a payment method on file). Credits cover Google Cloud Platform services and Firebase. Beyond credits, you get access to Google Cloud Skills Boost training, startup perks, and Google Workspace discounts.

---

## Eligibility Requirements

The Start tier has specific criteria. You must meet **all** of the following:

- **Technology startup** — your company must be building a technology product or service
- **Not yet funded by an institutional investor** — no venture capital, no accredited angel investors, no accelerator equity investment. Government grants, crowdfunding, prize funding, and friends & family funding are fine and will not disqualify you, but they also will not move you to the Scale tier
- **Founded within the last 5 years**
- **Have not previously received Google Cloud credits** beyond the standard $300 free trial
- **Not excluded** — Google will not accept companies that have IPO'd or been acquired, educational institutions, government entities, nonprofits, personal blogs/content sites, dev shops, consultancies, agencies, or cryptocurrency mining companies

Meeting all criteria does not guarantee approval. Google reviews every application individually.

---

## Registration (Step-by-Step)

1. **Create a Google Cloud account** — go to [console.cloud.google.com](https://console.cloud.google.com) and sign in with a Google account. If you already used the $300 free trial, you can use the same account
2. **Set up a Cloud Billing Account** — in the Google Cloud Console, go to Billing and create a billing account. Choose **Business** (not Individual) as the account type. You will need a credit card or bank account on file as a payment method
3. **Match your email domain** — the email on your Google Cloud Billing Account administrator must match your company website domain. For example, if your startup is at `acme.com`, use `you@acme.com` as the billing admin email. Personal Gmail addresses will not be accepted
4. **Have your company website ready** — you need a publicly accessible company website. The domain must match your billing email domain
5. **Note your Billing Account ID** — find your 18-character Billing Account ID in the Google Cloud Console under Billing. You will need this for the application
6. **Go to the application page** — visit [cloud.google.com/startup/apply](https://cloud.google.com/startup/apply)
7. **Switch to your business Google account** if prompted (not your personal Gmail)
8. **Fill out the application form** — provide details about your company, funding status (select pre-funded/bootstrapped), and cloud usage plans
9. **Submit and wait** — expect a response within **3-5 business days**. If your Billing Account is brand new, it may take **7-10 additional days** to process
10. **If approved** — credits are deposited directly to the Billing Account ID you provided. The 1-year validity clock starts immediately upon deposit

**Important caveats:**
- Apply with your **startup business email**, not a personal address
- If not approved, you will receive an email explaining why. Google's operations team may work with you to address gaps and requalify
- You can still use the $300 free trial while waiting for your application decision

---

## What the Credits Cover

The $2,000 in credits can be used for:

- **Google Cloud Platform services** — Compute Engine, Cloud Run, BigQuery, Cloud Storage, Cloud SQL, GKE, and all other standard GCP services
- **Firebase** — Firestore, Cloud Functions, Hosting, Authentication, and other Firebase services
- **Select Google Cloud offerings** — Looker and similar products
- **Vertex AI** — including Gemini API calls through Vertex AI, model training, and inference

**Not covered:**
- Google Cloud Marketplace third-party products and services
- Google Maps Platform (separate $600/month credit available for new signups)
- Google Workspace subscription fees (separate discount available)

---

## Additional Program Benefits

Beyond the $2,000 in cloud credits, the Start tier includes:

| Benefit | Details |
|---|---|
| **Google Cloud Skills Boost** | Training credits for cloud and AI courses |
| **Google Workspace** | Discounted pricing for business email and productivity tools |
| **Startup Perks** | Various partner offers and discounts |
| **Community Access** | Access to Google for Startups global network and events |
| **Technical Resources** | Documentation, guides, and startup-focused content |

---

## What Happens When Credits Expire

- Credits expire **1 year after deposit**, regardless of how much you have used
- Your running services are **not automatically shut down** — charges shift to your linked payment method (credit card or bank account)
- If you have no valid payment method or your account has no remaining credits, services may be suspended
- Set up **budget alerts** in the Billing console to avoid surprise charges when credits run out

---

## Upgrading to Scale Tier

If your startup raises equity funding (seed to Series A), you can reapply for the **Scale tier**, which offers significantly more:

| Tier | Credits | Duration |
|---|---|---|
| **Start** | Up to $2,000 | 1 year |
| **Scale** | Up to $200,000 | 2 years (Year 1: 100% up to $100K, Year 2: 20% up to $100K) |
| **Scale AI** | Up to $350,000 | 2 years (Year 1: 100% up to $250K, Year 2: 20% up to $100K) |

To upgrade, simply reapply at [cloud.google.com/startup/apply](https://cloud.google.com/startup/apply) once you have verifiable equity funding. Your existing Start credits will be incorporated. Stealth-mode companies with unannounced funding may take up to 2 weeks to verify.

---

## Additional Tips

- **Domain matching is strict** — the most common rejection reason is a mismatch between your billing email domain and your company website domain. Set this up correctly before applying
- **New billing accounts delay processing** — if you create a GCP billing account just before applying, expect the longer 7-10 day processing window. Consider setting up your account a few days before submitting
- **Combine with the free trial** — if you are a new GCP customer, you can also claim the separate $300 free trial. The free trial credits are used first before startup program credits
- **SAFE agreements count as equity for Scale** — if you raise via SAFE or convertible notes from accredited investors, you qualify for Scale, not just Start
- **Contact support** — if you have issues, email cloudstartupsupport@google.com
- **No GPU restriction** — unlike the $300 free trial, startup program credits on a paid billing account do not have GPU restrictions
- **Firebase included** — many founders miss that Firebase usage is covered by these credits, making it a good option for mobile app MVPs

---

**Sources:**
- [Google Cloud Startup Program](https://cloud.google.com/startup)
- [Google for Startups Cloud Program Application](https://cloud.google.com/startup/apply)
- [Program Eligibility and Benefits](https://cloud.google.com/startup/benefits)
- [Program FAQ](https://cloud.google.com/startup/faq)
- [Onboarding Checklist](https://cloud.google.com/startup/onboarding)
- [Google for Startups Cloud Overview](https://startup.google.com/cloud/)
- [Rho Guide to Google Cloud Credits](https://www.rho.co/blog/google-cloud-credits)
- [Start Tier (Pre-funded)](https://cloud.google.com/startup/pre-funded)
- [Scale Tier (Early Stage)](https://cloud.google.com/startup/early-stage)
- [Scale Tier AI](https://cloud.google.com/startup/ai)

## Submission 36: Google for Startups Cloud Program - Scale Tier ($200,000)

- Category: Startup Programs
- Source Link: https://cloud.google.com/startup/benefits
- Amount: Up to $200,000 (up to $350,000 for AI-first startups)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Equity-funded startups (pre-seed to Series A) can get up to **$200,000 in Google Cloud and Firebase credits** over two years through the Google for Startups Cloud Program Scale tier. **Year 1** covers **100% of usage up to $100,000**. **Year 2** reimburses **20% of the previous month's usage**, up to an additional **$100,000**. The program also includes **$12,000 in Enhanced Support credits** for one year, a dedicated **Startup Success Manager**, and access to Google Cloud engineers. AI-first startups may qualify for an expanded **$350,000 package** with additional Vertex AI resources.

---

## Eligibility Requirements

The Scale tier has strict eligibility criteria. You must meet **all** of the following:

- **Equity-funded** from **pre-seed to Series A** by an **institutional investor** (VC firm or equivalent). If Series A, it must have been raised within the **last 12 months**
- Founded within the **last 10 years**
- Have **not yet received more than $5,000** in Google Cloud credits (including the standard $300 free trial)
- Your company **website domain must match** your application email domain

**Qualifying funding types:**
- Equity investment from institutional investors and VC firms
- Simple Agreement for Future Equity (SAFE) agreements
- Verifiable token raises and blockchain foundation grants (Web3)

**NOT qualifying funding types:**
- Private equity
- Government innovation grants
- Prize funding or crowdfunding
- Angel investment
- Friends and family rounds

These non-qualifying funding types may still get you into the **Start tier** ($2,000 credits) but not Scale.

**Ineligible organizations:**
- Companies that have IPO'd or been acquired
- Educational institutions, government entities, nonprofits
- Personal blogs, content sites, dev shops, consultancies, or agencies

---

## Registration (Step-by-Step)

1. **Set up a Google Cloud account** at [console.cloud.google.com](https://console.cloud.google.com) if you don't have one. Create a billing account and note your **18-character billing account ID**
2. Go to the application page at [cloud.google.com/startup/apply](https://cloud.google.com/startup/apply)
3. **Use your company email** — it must match your startup's public website domain. Gmail or personal email will not work
4. Fill in company details: name, website URL, founding date, description of what you build
5. **Provide funding verification** — link to publicly available investment information (Crunchbase, PitchBook, Tracxn, investor websites, press releases). If in stealth mode, note that verification may take up to **two weeks**
6. Enter your **Google Cloud billing account ID** (18 characters). Google may auto-populate this from your email; verify it matches the account you want credits deposited into
7. List any **accelerator, incubator, or VC partner** affiliations — this strengthens your application and may unlock the AI tier
8. Submit and wait for review

**Timeline:**
- Standard review: **3-5 business days**
- New GCP billing accounts: may require **7-10 additional days**
- Stealth-mode companies (unannounced funding): up to **2 weeks** for funding verification

**Important:**
- Approval is at Google's discretion — meeting eligibility requirements does not guarantee acceptance
- If your business email domain differs from your website domain, you will need to prove ownership of both via your domain registrar

---

## Credit Structure (How It Actually Works)

### Year 1 — Full Coverage (up to $100,000)

Google covers **100%** of your Google Cloud and Firebase usage, up to $100,000. Credits are applied automatically to your billing account. If you spend less than $100,000 in Year 1, unused credits do **not** roll over.

### Year 2 — 20% Reimbursement (up to $100,000)

Year 2 works differently. Instead of a lump-sum credit, you receive credits equal to **20% of your previous month's usage**:

| Month | Your GCP Spend | Credits Received |
|---|---|---|
| December (Y1 ends) | $50,000 | -- |
| January (Y2 starts) | $60,000 | $10,000 (20% of Dec) |
| February | $70,000 | $12,000 (20% of Jan) |
| March | $80,000 | $14,000 (20% of Feb) |
| ... | ... | Continues until $100K cap or Y2 ends |

This means to maximize Year 2 credits, you need to be **spending at scale**. A startup spending $10,000/month would receive only $2,000/month back — it would take over four years to hit the $100K cap at that rate, but Year 2 only lasts 12 months.

---

## What Credits Cover (and What They Don't)

### Covered Services

| Category | Services |
|---|---|
| **Compute** | Compute Engine, GKE, App Engine, Cloud Run, Cloud Functions |
| **AI/ML** | Vertex AI, Vertex AI Studio, BigQuery ML |
| **Data** | BigQuery, Cloud SQL, Firestore, Bigtable, Cloud Spanner |
| **Storage** | Cloud Storage, Persistent Disks, Filestore |
| **Mobile/Web** | Firebase (all services) |
| **DevOps** | Cloud Build, Artifact Registry, Cloud Operations suite |
| **Analytics** | Looker (select offerings) |

### NOT Covered

- **Google Cloud Marketplace** third-party services and products
- **Google Workspace** — provided as a separate benefit (12 months free Business Plus for new signups)
- **Google Maps Platform** — separate benefit ($600/month for 12 months for new signups)
- Third-party partner model usage (except for AI tier startups who get a separate $10,000 for partner LLMs)

---

## Additional Benefits (Beyond Credits)

| Benefit | Details |
|---|---|
| **Enhanced Support** | $12,000 in credits for Google Cloud Enhanced Support for one year — higher-priority technical support |
| **Startup Success Manager** | Dedicated point of contact for guidance on architecture, cost optimization, and growth |
| **Google Cloud Engineers** | Access to technical experts for hands-on help |
| **Startup Community** | Global Google Cloud Startup Community access |
| **Co-marketing** | Select startups get co-marketing opportunities with Google |
| **Google Workspace** | 12 months of free Business Plus for new signups |
| **Google Maps Platform** | $600/month credits for 12 months (new signups) |
| **Upskill Opportunities** | Training and certification resources |

---

## AI Startup Bonus (Up to $350,000)

If your startup is **AI-first** — meaning AI is foundational to your product, not just a bolt-on feature — you may qualify for an expanded package:

- **Year 1:** 100% coverage up to **$250,000** (vs. $100,000 standard)
- **Year 2:** 20% reimbursement up to **$100,000** (same as standard)
- **Total:** Up to **$350,000** over 2 years
- **Extra:** $10,000 specifically for partner LLM models through Vertex AI Model Garden
- **Extra:** Dedicated AI training, webinars, and AI-specific resources

You apply through the same form. Being referred by a partner VC or accelerator significantly increases your chances of qualifying for the AI tier.

---

## Tips for Getting Approved

- **Apply through a VC or accelerator partner** if possible — this can significantly boost your credit allocation and approval chances. Ask your investors if they have a Google Cloud startup referral program
- **Have your funding publicly verifiable** — Crunchbase profiles, press releases, or investor website mentions speed up the process
- **Highlight Google Cloud usage** — explain how GCP fits into your tech stack and why you need the credits
- **Show traction** — metrics on users, revenue, or growth strengthen your application
- **Start with what you can get** — even if initially approved for a smaller amount, you can request an upgrade as your usage grows

---

## Gotchas and Caveats

- **Year 2 credits are retroactive**, not proactive — you pay full price and get 20% back the following month as credits. Cash flow planning matters
- **Credits are non-transferable and non-refundable** — you cannot sell, trade, or cash them out
- **Google has changed credit scope before** — some startups have reported that Google narrowed covered services mid-program without notice. Monitor your billing closely
- **No auto-renewal** — when the 2-year program ends, you move to standard pay-as-you-go pricing. Set budget alerts early
- **GPU restrictions may apply** during the standard free trial, but the startup program operates on a paid billing account so GPU access should be available
- **One application per startup** — if denied, you may reapply later but there is no guaranteed re-evaluation timeline

---

## Comparison with Other Tiers

| Feature | Start Tier | Scale Tier | AI Tier |
|---|---|---|---|
| **Credits** | $2,000 (1 year) | $200,000 (2 years) | $350,000 (2 years) |
| **Funding Required** | No | Yes (equity, pre-seed to Series A) | Yes (equity + AI-first) |
| **Year 1** | 100% up to $2K | 100% up to $100K | 100% up to $250K |
| **Year 2** | None | 20% up to $100K | 20% up to $100K |
| **Enhanced Support** | No | $12,000 credits | $12,000 credits |
| **Success Manager** | No | Yes | Yes |
| **Partner LLM Credits** | No | No | $10,000 |

---

## Additional Tips

- **Combine with the $300 free trial** — if you are new to Google Cloud, you can use the standard $300 free trial first and still apply for the startup program (as long as you haven't exceeded $5,000 in total credits)
- **Firebase counts** — Firebase usage is covered by startup credits, making this program valuable for mobile/web app startups
- **Vertex AI is fully covered** — unlike the free trial which restricts GPUs, the startup program runs on a paid account with full access to Vertex AI, Gemini models, and GPU instances
- **Google Workspace is separate** — don't expect cloud credits to cover your team's email and productivity tools; those are a separate 12-month benefit
- **Set budget alerts immediately** — Google Cloud has no spending caps by default. If you exceed your credits, you will be billed at standard rates

---

**Sources:**
- [Google for Startups Cloud Program - Benefits & Eligibility](https://cloud.google.com/startup/benefits)
- [Google for Startups Cloud Program - Apply](https://cloud.google.com/startup/apply)
- [Google for Startups Cloud Program - FAQ](https://cloud.google.com/startup/faq)
- [Google for Startups Cloud Program - Early Stage](https://cloud.google.com/startup/early-stage)
- [Google for Startups Cloud Program - AI Startups](https://cloud.google.com/startup/ai)
- [Google for Startups Overview](https://cloud.google.com/startup)
- [Google Cloud Startup Credits Guide - Vestbee](https://www.vestbee.com/insights/articles/google-cloud-credits-for-startups-a-comprehensive-guide)
- [Google Cloud Credits for Startups - Rho](https://www.rho.co/blog/google-cloud-credits)
- [2026 Ultimate Guide to Google for Startups - DEV Community](https://dev.to/googleai/the-2026-ultimate-guide-to-google-for-startups-funding-accelerators-and-350k-in-credits-10ha)

## Submission 35: Google for Startups Cloud Program - AI Track ($350,000)

- Category: Startup Programs
- Source Link: https://cloud.google.com/startup/ai
- Amount: Up to $350,000 in Google Cloud credits over 2 years, plus $12,000 in support credits and $10,000 in partner model credits
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

AI-first startups building on Vertex AI or Gemini can receive **up to $350,000 in Google Cloud credits over 2 years** through the AI Track of the Google for Startups Cloud Program. Year 1 covers **100% of eligible Google Cloud usage up to $250,000**. Year 2 covers **20% of usage up to an additional $100,000**. On top of the cloud credits, you get **$12,000 in Enhanced Support credits**, **$10,000 in partner model credits** (Anthropic, Fireworks AI, and others via Vertex AI Model Garden), a dedicated Startup Success Manager, and access to DeepMind mentorship. You must have equity funding from seed to Series A and demonstrate that Vertex AI or Gemini is core to your product.

---

## Registration (Step-by-Step)

1. **Create a Google Cloud account** at [console.cloud.google.com](https://console.cloud.google.com) if you don't have one
2. **Set up a Cloud Billing Account** in the Google Cloud Console — navigate to Billing and create a new billing account. Note your **18-character billing account ID** (format: ABC123-DEF456-GHI789)
3. **Wait 48 hours** after creating your billing account before applying — Google needs time to sync the ID to their systems
4. **Prepare your application materials:**
   - Your 18-character Google Cloud billing account ID
   - A business email that matches your startup's public website domain (e.g., yourname@yourstartup.com, not a Gmail address)
   - Links to publicly available sources documenting your equity funding (e.g., Crunchbase profile, press releases, investor announcements)
   - If you're in stealth mode, coordinate with your investor to verify your funding directly
5. Go to [cloud.google.com/startup/apply](https://cloud.google.com/startup/apply) and complete the application form
6. Select the **AI Track** and describe how Vertex AI or Gemini is central to your product
7. Submit and wait — expect a response within **3-5 business days**. New billing accounts may take 7-10 additional days. Stealth-mode startups may wait up to 2 weeks for funding verification
8. If approved, credits are deposited immediately into your billing account

**Important:**
- Your company website domain **must match** your email domain and the email domain on your billing account
- Google auto-populates your billing ID based on the email you provide — double-check it matches the account you want credits deposited into
- You cannot have previously received more than $5,000 in Google Cloud credits
- One enrollment per startup — if denied, Google's operations team will explain why and may help you requalify

---

## Credit Structure

| Period | Coverage | Maximum |
|---|---|---|
| **Year 1** | 100% of eligible Google Cloud usage | $250,000 |
| **Year 2** | 20% of eligible Google Cloud usage | $100,000 |
| **Enhanced Support** | Google Cloud Enhanced Support | $12,000 (1 year) |
| **Partner Models** | Vertex AI Model Garden partner LLMs | $10,000 |
| **Total** | | **Up to $372,000** |

The Year 1 amount includes the standard Scale Tier credits ($100,000) plus an additional $150,000 for AI Tier enrollment. Year 2 is a 20% reimbursement model — you pay 80% and credits cover the remaining 20%, up to $100,000.

---

## Eligibility Requirements

**Who qualifies:**
- AI-first startups using or planning to use **Vertex AI or Gemini** as a foundation of their primary product
- Received **equity funding from seed to Series A** (if Series A, must have been raised within the last 12 months)
- Company must be **under 10 years old**
- Must not have previously received more than $5,000 in Google Cloud credits

**Accepted funding types:**
- Equity investments from institutional investors and venture capital firms
- Simple Agreement for Future Equity (SAFE) agreements and convertible notes
- Verifiable token raises and blockchain foundation grants (Web3 community)

**Not accepted for Scale/AI tier (evaluated for Start tier instead):**
- Private equity funding
- Government innovation grants
- Prize funding
- Crowdfunding
- Angel investors only (without institutional backing)
- Friends and family funding

**Excluded organizations:**
- Companies that have IPO'd or been acquired
- Educational institutions, government entities, nonprofits
- Personal blogs, dev shops, consultancies, agencies
- Cryptocurrency mining companies
- Companies distributing tokens contrary to regulatory guidance

---

## What Credits Cover (and Don't Cover)

**Credits CAN be used for:**
- Google Cloud Platform services (Compute Engine, GKE, Cloud Run, BigQuery, etc.)
- Firebase services
- Vertex AI (including Gemini model API calls)
- Select Google Cloud offerings like Looker

**Credits CANNOT be used for:**
- Third-party services on Google Cloud Marketplace
- Google Workspace (separate benefit — 12 months of Business Plus free for new signups)
- Google Maps Platform (separate benefit — $600/month for 12 months for new signups)
- Any non-Google Cloud services

Credits are non-cash, non-transferable, and apply only to eligible service SKUs. Usage beyond credit coverage bills at standard rates.

---

## Partner Perks & Model Garden Credits

AI Tier startups receive **$10,000 in credits for partner LLM models** accessible through Vertex AI Model Garden. This is separate from the main cloud credits and covers:

| Partner | Benefit |
|---|---|
| **Anthropic (Claude)** | Up to $10,000 in model credits via Model Garden |
| **Fireworks AI** | $10,000 in serverless inference credits (hundreds of open-source models) |
| **ElevenLabs** | 33 million free speech synthesis credits |
| **MongoDB Atlas** | Startup database credits |
| **CockroachDB** | Distributed SQL support |
| **GitLab Ultimate** | 1 full year free |
| **Contextual AI** | $1,500 in credits |

To activate partner model credits, contact your Google Cloud Account Executive or Startup Success Manager after enrollment.

---

## Mentorship & Support

- **Startup Success Manager** — dedicated Google Cloud contact for technical and business guidance
- **DeepMind mentorship** — research partnership opportunities with teams building Gemini and other frontier models
- **$500 in Google Cloud Skills Boost** training credits (Scale tier gets $200)
- **Frontier model access** — early testing of Gemini, Veo, and experimental features
- **AI Startup School** — Google-hosted events for AI founders
- **Google Cloud Next** — priority access to Google's annual conference
- **AI/ML architecture resources** and generative AI training courses

---

## Program Tiers Comparison

| Tier | Credits | Target | Funding Required |
|---|---|---|---|
| **Start** | $2,000 (1 year) | Pre-funded startups | None |
| **Scale** | Up to $200,000 (2 years) | Funded startups | Seed to Series A |
| **Scale AI** | Up to $350,000 (2 years) | AI-first startups | Seed to Series A |
| **Web3** | Specialized | Blockchain startups | Varies |

If you don't qualify for the AI Track, you may still qualify for the Scale tier ($200,000) or the Start tier ($2,000).

---

## Tips for Getting Approved

- **Get a VC or accelerator referral** — having a partner recommend you significantly increases approval odds, especially for the AI Track
- **Prepare a one-pager** covering your problem, solution, metrics, forecasted cloud spend, and reference architecture on Google Cloud
- **Emphasize Vertex AI / Gemini usage** — clearly explain how Google's AI services are core to your product, not just peripheral
- **Match your domains** — your email, company website, and billing account must use the same domain
- **Apply early in your funding journey** — the pre-seed to Series A window is the sweet spot
- **Don't wait to apply** — the program accepts rolling applications year-round, no batch deadlines
- **Set budget alerts** after approval — Google Cloud has no automatic spending caps, and usage beyond your credits bills at standard rates

---

## What Happens When Credits Expire

- After the 2-year program period, all credits expire and you transition to standard pay-as-you-go billing
- Always-free tier services (Cloud Vision, Speech-to-Text, Natural Language, etc.) continue indefinitely within their monthly limits
- Set up billing alerts well before expiration to avoid surprise charges
- You can continue using all Google Cloud services at standard pricing

---

**Sources:**
- [Google for Startups Cloud Program - AI Track](https://cloud.google.com/startup/ai)
- [Google for Startups Cloud Program - Apply](https://cloud.google.com/startup/apply)
- [Program Eligibility and Benefits](https://cloud.google.com/startup/benefits)
- [Program FAQs](https://cloud.google.com/startup/faq)
- [Startup Perks](https://cloud.google.com/startup/perks)
- [Google for Startups - Cloud](https://startup.google.com/cloud/)
- [2026 Ultimate Guide to Google for Startups (DEV Community)](https://dev.to/googleai/the-2026-ultimate-guide-to-google-for-startups-funding-accelerators-and-350k-in-credits-10ha)
- [How to Get Up to $350,000 in Cloud Credits (Mobitouch)](https://mobitouch.net/blog/how-to-get-up-to-350000-in-cloud-credits-for-startups-2025-guide)

## Submission 34: Google Colab Free Tier - T4 GPU Access

- Category: Free GPU & Compute
- Source Link: https://colab.research.google.com/
- Amount: Free T4 GPU access with dynamic usage limits; ~12-hour max sessions, ~15-30 GPU hours/week
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Google Colab gives you a **free browser-based Jupyter notebook** with access to an **NVIDIA T4 GPU** (16 GB VRAM) and optional **TPU** runtime -- no credit card, no signup form, no approval queue. Just sign in with any Google account and start running Python. Sessions last up to **12 hours**, you get roughly **12-13 GB of system RAM**, and you can run **2 notebooks concurrently**. GPU access is not guaranteed during peak usage, but is usually available. This is the fastest path from zero to running inference on a 7B-parameter LLM, fine-tuning BERT, generating images with Stable Diffusion, or transcribing audio with Whisper.

---

## Registration (Step-by-Step)

1. Go to [colab.research.google.com](https://colab.research.google.com)
2. Sign in with **any Google account** (personal Gmail works fine)
3. Click **"New Notebook"** to create a blank notebook
4. To enable GPU: go to **Runtime > Change runtime type > Hardware accelerator > T4 GPU** (or TPU)
5. Click **Connect** in the top-right corner -- Colab provisions a VM with the selected accelerator
6. Start writing and running Python code immediately

**Important:**
- No credit card or payment method required -- ever
- No separate signup or application process; your Google account is all you need
- Notebooks are saved to your **Google Drive** by default (you can also load from GitHub)
- All data on the VM is **ephemeral** -- files saved locally on the VM vanish when the session ends. Mount Google Drive or download files to persist them

---

## GPU & TPU Hardware

| Accelerator | Free Tier Availability | VRAM / Memory | Notes |
|---|---|---|---|
| **NVIDIA T4** | Default GPU assignment | 16 GB (15 GB usable after ECC) | Turing architecture; good for inference and small-to-medium training |
| **TPU v2** | Available (select in runtime settings) | 8 GB HBM per core | Best for JAX/TensorFlow transformer workloads |

- GPU/TPU type and availability **vary over time** and are **not guaranteed**. During peak hours you may only get CPU
- The T4 has **320 GB/s memory bandwidth**, 70W power draw, and Tensor Cores for mixed-precision (FP16) acceleration
- Paid tiers unlock faster GPUs (V100, A100) with priority access

---

## Resource Limits

| Resource | Free Tier Limit |
|---|---|
| **System RAM** | ~12-13 GB |
| **GPU VRAM** | 16 GB (T4) |
| **Disk space** | ~35-78 GB ephemeral (varies per session) |
| **Max session duration** | 12 hours |
| **Idle timeout** | ~90 minutes of no interaction |
| **Concurrent notebooks** | 2 |
| **Weekly GPU hours** | ~15-30 hours (dynamic, not officially published) |
| **Terminal access** | Not available (Pro+ only) |
| **Background execution** | Not available (Pro+ only) |

Google intentionally does not publish exact quotas -- they fluctuate based on overall demand and your recent usage patterns. If you exhaust your GPU quota for the period, you can still use CPU-only runtimes.

---

## What You Can Run on the Free T4

The T4 with 16 GB VRAM is surprisingly capable for a free resource. Here is a realistic breakdown:

**Works well:**
- Inference on **quantized 7B-parameter LLMs** (Llama 2 7B, Mistral 7B with 4-bit quantization via bitsandbytes or GPTQ)
- Fine-tuning **BERT, GPT-2, DistilBERT** and similar encoder/decoder models
- **Stable Diffusion** image generation (SD 1.5, SDXL with optimizations)
- **Whisper** speech-to-text (all model sizes up to large-v3)
- Training **CNNs** for image classification (ResNet, EfficientNet)
- Running **LangChain / RAG** pipelines with local embedding models
- **LoRA / QLoRA fine-tuning** of 7B models with small batch sizes

**Tight but possible with tricks:**
- Fine-tuning 13B models with QLoRA, gradient checkpointing, and batch size 1
- Stable Diffusion XL without memory optimizations

**Will not fit:**
- Training or full-precision inference on 30B+ parameter models
- Large batch sizes with big sequence lengths (e.g., batch=4, seq_len=2048 on a 7B model)
- Any workload requiring more than 16 GB VRAM

**Tip:** Enable **mixed-precision training** (FP16) to take advantage of the T4's Tensor Cores -- this can cut memory usage and training time significantly.

---

## Session Management & Idle Timeout

- **90-minute idle timeout:** If you do not interact with the notebook (click, type, scroll) for about 90 minutes, Colab disconnects your runtime and reclaims the VM. Interaction means activity in the browser tab -- code running in the background does not count as interaction
- **12-hour hard cap:** Even with continuous interaction, the session terminates after 12 hours. Plan your training runs accordingly and checkpoint frequently
- **Reconnection:** If disconnected, you can reconnect to a new runtime, but all local files and variables are lost. Mount Google Drive and save checkpoints there
- **Quota cooldown:** If you use a lot of GPU time in a short period, Colab may temporarily restrict you to CPU-only runtimes. Waiting a few hours or until the next day usually restores GPU access

---

## Storage & File Persistence

- The VM disk (35-78 GB) is **ephemeral** -- everything is wiped when the session ends
- **Google Drive integration:** Run `from google.colab import drive; drive.mount('/content/drive')` to mount your Drive. Files saved there persist across sessions
- Google Drive storage quota is separate from the VM disk. Free Google accounts get **15 GB** of Drive storage (shared with Gmail and Photos)
- You can also upload/download files directly, clone Git repos, or use `gdown` for Google Drive links

---

## Prohibited Uses

Google explicitly prohibits using free Colab for:
- Cryptocurrency mining
- Running web servers, proxies, or file hosting
- SSH/remote desktop access
- Torrenting or peer-to-peer file sharing
- Password cracking or denial-of-service attacks
- Deepfake generation that bypasses policies
- Bypassing the notebook UI for automated content generation

Violating these policies can result in temporary or permanent restriction of your Colab access.

---

## Paid Tier Comparison

| Feature | Free | Pro ($9.99/mo) | Pro+ ($49.99/mo) | Pay-as-you-go |
|---|---|---|---|---|
| **GPU** | T4 (not guaranteed) | Priority T4, V100, A100 | Premium A100 | Based on compute units |
| **RAM** | ~12-13 GB | Up to 32 GB | Up to 52 GB | Varies |
| **Max session** | 12 hours | 24 hours | 24 hours | 24 hours |
| **Concurrent notebooks** | 2 | More | More | More |
| **Terminal** | No | No | Yes | No |
| **Background execution** | No | No | Yes | No |
| **Compute units** | N/A | 100/month | 500/month | 100 for $9.99 |

A T4 consumes about **11.7 compute units per hour**; an A100 consumes about **62 CU/hr**. When paid users exhaust their compute units, they revert to free-tier policies.

---

## Additional Tips

- **Kaggle Notebooks** offer a comparable free alternative: 30 hours/week of GPU (NVIDIA P100, 16 GB VRAM) and free TPU v3-8 access. Worth using as a supplement when Colab GPU quota is exhausted
- **Google AI Studio** (aistudio.google.com) provides free Gemini API access with no credit card -- a separate free resource for LLM API calls
- **Checkpoint frequently** -- save model weights to Google Drive every N epochs. Session disconnects are inevitable
- **Use `!nvidia-smi`** in a code cell to verify your GPU type and check VRAM usage
- **Mixed precision** (`torch.float16` or `torch.bfloat16`) and **gradient checkpointing** can reduce VRAM usage by 40-60%
- **Avoid leaving idle tabs open** -- Colab may throttle users who habitually hold GPU sessions without active use
- **Colab Enterprise** is a separate, paid Google Cloud product with different pricing and is not related to the free Colab tier

---

**Sources:**
- [Google Colab](https://colab.research.google.com)
- [Google Colab FAQ](https://research.google.com/colaboratory/faq.html)
- [Colab Paid Services Pricing](https://colab.research.google.com/signup)
- [Understanding Google Colab Free GPU in Detail](https://medium.com/data-science-in-your-pocket/understanding-google-colab-free-gpu-in-detail-15074081d494)
- [The Complete Guide to Google Colab for Free AI Development](https://www.aifire.co/p/the-complete-guide-to-google-colab-for-free-ai-development)
- [Free Cloud GPU Comparison 2026](https://research.aimultiple.com/free-cloud-gpu/)
- [Colab GPUs Features & Pricing](http://mccormickml.com/2024/04/23/colab-gpus-features-and-pricing/)

## Submission 33: Google Cloud Teaching & Learning Credits

- Category: Education & Student
- Source Link: https://cloud.google.com/edu/students
- Amount: $50 per student per course; $100 per faculty member per course
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Faculty at accredited higher education institutions can apply for **free Google Cloud credits** to distribute to students enrolled in their courses. Students receive **$50 in GCP credits** (faculty get **$100**) via coupon codes redeemable at the Google Cloud Console. **No credit card required.** Credits cover virtually all GCP services -- Compute Engine, BigQuery, Vertex AI, Cloud Storage, Vision API, Translation API, and more -- for course-related projects and assignments. The program spans **75+ countries** and is designed for hands-on cloud learning in the classroom.

---

## How It Works (Overview)

This is **not a self-serve student program**. The flow is:

1. **Faculty applies** for teaching credits through the Google for Education portal
2. Google reviews and approves the application (up to **15 business days**)
3. Faculty receives a **URL** to distribute to students
4. Students visit the URL, verify with their **school email**, and receive a **coupon code**
5. Students redeem the coupon at the Cloud Console -- a billing account with credit is created automatically

If you are a student and your professor has not applied, you cannot get these credits on your own. Ask your faculty to apply.

---

## Registration — Faculty (Step-by-Step)

1. Go to [edu.google.com/programs/credits/teaching](https://edu.google.com/intl/ALL_us/programs/credits/teaching/)
2. Click **"Apply now"**
3. Fill in **personal details**: name, official school email (must be on the institution's domain, e.g. `.edu`), country, state
4. Provide **institution and department** information
5. Add a **link to your faculty directory profile** (Google uses this to verify your role)
6. Enter **course details**: course name, start date, abstract, course URL. You can request credits for up to **3 courses** per application
7. Specify the **expected number of students and TAs/staff**
8. Submit the application
9. Wait for approval -- processing takes up to **15 business days**
10. Upon approval, you receive **two emails**: one with your instructor credits ($100) and one with a **student distribution URL**

**Important:**
- Applications are accepted on a rolling basis
- Google will not approve courses starting more than **1 year** from the application date
- Only the most recent application submission is reviewed (resubmitting overwrites previous)
- For-profit institutions are **not eligible**
- Only faculty can apply -- students cannot apply directly

---

## Registration — Student (Step-by-Step)

1. Your faculty member sends you a **URL** (from the approval email) to request your coupon
2. Visit the URL and enter your **school-branded email address** (e.g. `yourname@university.edu`)
3. Verify your email -- you will receive a **coupon code** via email
4. Go to [console.cloud.google.com/education](https://console.cloud.google.com/education)
5. Sign in with your **Google account** (can be personal Gmail or Google Workspace for Education)
6. The coupon code will be **pre-populated** -- click **"Accept and Continue"**
7. A new **Cloud Billing account** is created with your $50 credit loaded
8. Create or link a **GCP project** to this billing account to start using services

**Important:**
- **No credit card required** -- if you are prompted for credit card info, you are accidentally using the Free Trial instead of education credits
- One coupon per student per course
- Up to **3 coupon codes** can be redeemed per course (if faculty provides additional coupons)
- Each course gets its own **separate billing account** for tracking
- Once applied to an account, the coupon **cannot be transferred** to a different account

---

## Covered GCP Services

Education credits can be used for **virtually all Google Cloud services**, including:

| Category | Services |
|---|---|
| **Compute** | Compute Engine (VMs), Cloud Run, App Engine, Cloud Functions |
| **AI / ML** | Vertex AI, AutoML, Vision API, Natural Language API, Translation API, Speech-to-Text, Text-to-Speech |
| **Data** | BigQuery, Dataflow, Dataproc, Pub/Sub |
| **Storage** | Cloud Storage, Firestore, Cloud SQL, Bigtable |
| **Networking** | VPC, Cloud Load Balancing, Cloud CDN |
| **Maps** | Maps Platform products (Maps, Routes, Places APIs) |
| **Containers** | Google Kubernetes Engine (GKE), Artifact Registry |

Services in the **GCP Free Tier** (e.g. first 1 TB of BigQuery queries/month) do **not** consume your education credits -- those free allowances apply on top.

---

## GPU Access

GPUs are **not enabled by default** on education credit accounts, but you can request access:

1. In the Cloud Console, go to **IAM & Admin > Quotas**
2. Filter by **"GPUs (all regions)"**
3. Select the quota, click **"Edit Quotas"**, and request the desired limit
4. Submit the request -- approval can take **minutes to a week**

Unlike the $300 Free Trial (which blocks GPU access entirely), education credit accounts **can** request GPU quota. However, approval is not guaranteed and depends on availability. Some zones may not have certain GPU types. Check [GPU pricing and availability](https://cloud.google.com/compute/gpus-pricing) before requesting.

---

## Credit Amounts & Expiration

| Recipient | Credit Amount |
|---|---|
| **Student** | Up to **$50** per course |
| **Faculty / Teaching Staff** | Up to **$100** per course |

- **Coupon redemption window:** 16 weeks from the course start date indicated on the application
- **Credit validity:** 12 months from redemption
- Faculty can request **additional credits** by emailing **CloudEduGrants@google.com**

---

## When Credits Run Out

- Running VMs are **stopped** (not deleted)
- Storage (including VM disks) persists for **30 days**, then is **permanently deleted**
- You can no longer use paid GCP services, but **Free Tier** services continue to work
- Contact your faculty member to request additional coupons

---

## Eligible Countries

The program is available in **75+ countries**, including:

- **North America:** United States, Canada
- **Europe:** Austria, Belgium, Bulgaria, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Spain, Sweden, Switzerland, United Kingdom
- **Middle East / Africa:** Egypt, Ghana, Israel, Kenya, Morocco, Namibia, Nigeria, Senegal, South Africa, Tunisia
- **Asia-Pacific:** Australia, Hong Kong, Indonesia, Japan, Macau, Philippines, South Korea, Taiwan, Thailand, Vietnam
- **South America:** Peru, Uruguay

In the US, institutions must be **regionally accredited**. Outside the US, institutions must be **degree-granting universities or higher learning institutions**. For-profit institutions are **not supported**. Countries not on the list can express interest for future inclusion.

---

## Restrictions

- Credits are for **educational use only** -- no commercial or personal projects
- **Cryptocurrency mining** is explicitly prohibited
- Credits are **non-transferable** between accounts
- Credits are awarded at **Google's discretion** and can be revoked
- Only school-domain email addresses can receive coupon codes
- Faculty must apply -- students cannot self-enroll

---

## Related Google Cloud Education Programs

| Program | Who | Amount |
|---|---|---|
| **Research Credits** | Faculty / Postdocs | Up to **$5,000** |
| **Research Credits** | PhD Students | Up to **$1,000** |
| **Training Credits** (Skills Boost) | Faculty | Up to **5,000** Google Cloud Skills Boost credits to distribute |
| **Training Credits** (Skills Boost) | Students / IT Admins | **200** Google Cloud Skills Boost credits |
| **Google Cloud Free Trial** | Anyone (new customers) | **$300** for 91 days (credit card required) |
| **Gemini Advanced for Students** (US only) | Verified college students | Free access to Gemini Advanced, NotebookLM Plus, Whisk + 2 TB storage until finals 2026 (sign up by June 30, 2025) |

---

## Additional Tips

- **Faculty: redeem your own credits immediately** after approval -- Google recommends this to avoid issues
- **$50 goes fast** if you spin up VMs or use GPUs. Stick to lightweight instances and shut down resources when not in use
- **BigQuery Free Tier** gives you 1 TB of queries and 10 GB storage per month without touching credits -- use it
- **Vertex AI can be expensive** -- monitor usage carefully; a few hours of model training can consume most of your credits
- If you need **more credits**, have your faculty email **CloudEduGrants@google.com** -- additional allocations are possible
- **Do not confuse this with the Free Trial** -- the Free Trial requires a credit card and gives $300 for 91 days. Education credits have no credit card requirement and last 12 months
- **Set budget alerts** in the Billing console to get notified before credits run out

---

**Sources:**
- [Google Cloud Teaching Credits Application](https://edu.google.com/intl/ALL_us/programs/credits/teaching/)
- [Google Cloud for Education Overview](https://cloud.google.com/edu)
- [Google Cloud for Education - Students](https://cloud.google.com/edu/students)
- [Get and Redeem Education Credits (Cloud Billing Docs)](https://docs.cloud.google.com/billing/docs/how-to/edu-grants)
- [Credits and Coupon Administration](https://support.google.com/google-cloud-higher-ed/answer/10322773?hl=en)
- [About the Program](https://support.google.com/google-cloud-higher-ed/answer/10324788?hl=en)
- [Program Information](https://support.google.com/google-cloud-higher-ed/answer/10723679?hl=en)
- [Eligibility](https://support.google.com/google-cloud-higher-ed/answer/10322652?hl=en)
- [Faculty Managing Student Credits](https://support.google.com/google-cloud-higher-ed/answer/10324789?hl=en)
- [Google Cloud Expands Credits to 75+ Countries](https://cloud.google.com/blog/topics/public-sector/google-cloud-expands-higher-education-credits-support-75-countries-globally/)

## Submission 32: Google Cloud Research Credits - Up to $5,000

- Category: Education & Student
- Source Link: https://edu.google.com/intl/ALL_us/programs/credits/research
- Amount: Up to $1,000 (PhD students, renewable annually) or $5,000 (faculty/postdocs/nonprofit researchers, one-time)
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Google Cloud offers **free cloud credits for academic researchers** at accredited institutions worldwide. PhD students can receive **up to $1,000** (renewable annually), while faculty, postdocs, and nonprofit researchers can receive **up to $5,000** (one-time). Credits cover virtually any GCP service — Compute Engine, Vertex AI, BigQuery, Cloud Storage, GPU instances for ML training — with a few exclusions. No credit card is required; you apply with a research proposal and Google reviews it within 4-8 weeks.

---

## Who Qualifies (and Who Doesn't)

**Eligible:**
- **Faculty** at accredited higher education institutions
- **Postdoctoral researchers** at accredited institutions
- **PhD students** conducting research at accredited institutions
- **Nonprofit researchers** at eligible nonprofit research institutions whose mission aligns with expanding academic research and advancing open scientific knowledge

**Not eligible:**
- Master's degree students
- Graduate students who are not in a PhD program
- Undergraduate students
- Industry researchers or employees at for-profit companies
- Individuals at institutions in non-approved countries

Your institution must be regionally accredited. The program is available in **75+ countries** including the US, Canada, UK, Australia, and many countries across Europe, Asia, Africa, and Latin America. If your country or institution is not listed, you can submit an Express Interest form for future consideration.

---

## Credit Amounts & Renewal

| Role | Credit Amount | Renewal |
|---|---|---|
| **Faculty** | Up to $5,000 | One-time only |
| **Postdoctoral researcher** | Up to $5,000 | One-time only |
| **Nonprofit lab researcher** | Up to $5,000 | One-time only |
| **PhD student** | Up to $1,000 | Once per year, renewable across up to 5 years of the program |

Only **one person per research project** may apply. You cannot submit multiple applications for the same project to get more credits.

---

## Registration (Step-by-Step)

1. **Set up a Google Cloud billing account** before applying — you need a billing account ID for the application form. Go to [console.cloud.google.com/billing](https://console.cloud.google.com/billing) and create a billing account. No credit card is required for the credits themselves, but Google may require billing account verification
2. **Estimate your costs** using the [Google Cloud Pricing Calculator](https://cloud.google.com/products/calculator). You will need to paste the pricing calculator results URL into the application
3. Go to the application form at [edu.google.com/programs/credits/research](https://edu.google.com/intl/ALL_us/programs/credits/research/)
4. Fill in **personal information** — full name, institutional email (preferred), country
5. Fill in **organization details** — select your job title (Faculty, Postdoctoral Researcher, PhD Student), enter your institution name (autocomplete from institutional directory), department, and a link to your faculty/researcher directory profile
6. Write your **research proposal** (max 250 words) — describe the research problem, which GCP tools you plan to use, timeline, milestones, and expected outcomes
7. Select your **field of research** category
8. Provide a link to your **Google Scholar profile** (optional but recommended)
9. Enter your **Google Cloud billing account ID** and the pricing calculator results URL
10. Describe your **post-credit funding strategy** (what happens when credits run out)
11. Submit the application and wait for review

**Important:**
- Applications are accepted on a **rolling basis** — there is no deadline
- Google typically reviews applications within **4-8 weeks**
- If you have an urgent timeline, mention it in your proposal
- Credits are awarded at Google's discretion — not all applications are approved
- Confirmation emails may land in spam, so check your spam folder
- Contact **gcpresearchcredits@google.com** with questions

---

## What You Can Use Credits For

Research credits work like regular GCP credits and cover **most Google Cloud services**, including:

- **Compute Engine** — VMs, including GPU-attached instances (A100, T4, L4, etc.) for ML training
- **Vertex AI** — model training, tuning, deployment, Vertex AI Studio, Vertex AI Notebooks
- **BigQuery** — large-scale data analysis
- **Cloud Storage** — storing datasets, model artifacts, results
- **Cloud Run / Cloud Functions** — serverless compute
- **Dataflow / Dataproc** — batch and stream processing, Spark/Hadoop clusters
- **Cloud TPUs** — tensor processing units for ML workloads
- **AI/ML APIs** — Natural Language, Vision, Speech-to-Text, Translation, Video Intelligence
- **Kubernetes Engine** — container orchestration

---

## What You Cannot Use Credits For

Credits are **excluded** from the following:

| Excluded Service | Notes |
|---|---|
| **Google Maps Platform** | All Maps, Routes, and Places APIs |
| **Google Cloud Support packages** | Premium, Enhanced, or Standard support |
| **Google Colab products** | Colab Pro, Colab Pro+, Colab Enterprise |
| **Cryptocurrency mining** | Explicitly prohibited |
| **Commercial use** | Credits are for academic research only, not commercial products |
| **Personal benefit** | Cannot be used for personal, non-research purposes |

Credits are also **non-transferable** and may not be sold or bartered.

---

## GPU & ML Training Access

Unlike the standard $300 Free Trial (which blocks GPU usage), research credits **do allow GPU instances**. You can attach GPUs to Compute Engine VMs or use them through Vertex AI for ML training. Common GPU types available on GCP:

| GPU | Use Case | Approximate Cost/Hour |
|---|---|---|
| **NVIDIA T4** | Inference, light training | ~$0.35/hr |
| **NVIDIA L4** | Inference, medium training | ~$0.70/hr |
| **NVIDIA A100 (40GB)** | Heavy training, large models | ~$3.67/hr |
| **NVIDIA A100 (80GB)** | Large-scale training | ~$4.30/hr |
| **NVIDIA H100** | Frontier model training | ~$11.50/hr |

With $5,000 in credits, a faculty member could get roughly **1,350 hours of T4 GPU time** or **1,150 hours of A100 (40GB) time** — enough for significant ML research. PhD students with $1,000 get proportionally less, so plan your compute budget carefully using the pricing calculator.

**Tip:** Use preemptible/spot VMs (60-90% cheaper) to stretch your credits further. Spot T4 instances cost around $0.11/hr.

---

## Credit Activation & Expiration

- Credits must be **activated within 60 days** of the project start date you specified in your application
- Once activated, credits **expire after 365 days** or when fully used, whichever comes first
- Credits are added to a dedicated Cloud Billing account
- You can use GCP always-free tier services **without consuming** your research credits
- After credits expire, any running workloads will be stopped unless you have another funding source linked

---

## Writing a Strong Proposal

Your 250-word proposal is the most important part of the application. Google recommends including:

- **Clear research problem** — what question are you investigating?
- **Specific GCP services** — name the exact services you plan to use (e.g., "Vertex AI for fine-tuning a transformer model on our genomics dataset using A100 GPUs")
- **Timeline and milestones** — show you have a concrete plan
- **Expected outcomes** — publications, open-source tools, datasets
- **Cost justification** — tie your pricing calculator estimate to specific workloads

Vague proposals like "we want to explore cloud computing" are less likely to be approved than specific ones detailing exact workloads and resource needs.

---

## Referral Program

Google offers a **referral bonus** for existing credit recipients. If you refer **two qualified faculty researchers** who are approved for credits, you become eligible for an additional research grant. This is a useful way to extend your funding if you have colleagues who could benefit from the program.

---

## Additional Tips

- **No credit card required** for the credits themselves, but you need a GCP billing account set up before applying
- **PhD students can reapply annually** — plan your research in yearly increments to maximize total credits over your program
- **Combine with always-free tier** — services like BigQuery (1 TB queries/month), Cloud Storage (5 GB), and e2-micro VMs are always free and do not consume your credits
- **Set budget alerts** — configure billing alerts in the GCP console to avoid accidentally exceeding your credit balance
- **Use Spot VMs for training** — preemptible instances are 60-90% cheaper and work well for fault-tolerant ML training jobs
- **Google for Startups** — if your research becomes a startup, the [Google for Startups Cloud Program](https://cloud.google.com/startup) offers up to $200K in credits
- **Teaching Credits** — separate from research credits, faculty can also apply for teaching credits ($50-$100/student/semester) through the [Teaching Credits program](https://edu.google.com/intl/ALL_us/programs/credits/teaching/)

---

**Sources:**
- [Google Cloud Research Credits Program](https://edu.google.com/intl/ALL_us/programs/credits/research/)
- [Google Cloud for Researchers](https://cloud.google.com/edu/researchers)
- [Program Information - Cloud for Education Help](https://support.google.com/google-cloud-higher-ed/answer/10723679?hl=en)
- [Eligibility - Cloud for Education Help](https://support.google.com/google-cloud-higher-ed/answer/10324705?hl=en)
- [Application Guidelines - Cloud for Education Help](https://support.google.com/google-cloud-higher-ed/answer/10724468?hl=en)
- [University of Colorado - Apply for Google Cloud Research Credits](https://curc.readthedocs.io/en/latest/cloud/gcp/Google-Cloud-research-credits.html)
- [Google Cloud Expands Higher Education Credits to 75+ Countries](https://cloud.google.com/blog/topics/public-sector/google-cloud-expands-higher-education-credits-support-75-countries-globally/)
- [Get and Redeem Education Credits - Google Cloud Billing](https://docs.cloud.google.com/billing/docs/how-to/edu-grants)

## Submission 31: Google Cloud Free Trial - $300 Credits

- Category: Cloud Providers
- Source Link: https://cloud.google.com/free
- Amount: $300
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

New Google Cloud customers get **$300 in free credits valid for 91 days** to explore virtually all GCP services — including Vertex AI, Gemini API, Compute Engine, BigQuery, and Cloud Storage. Your billing account auto-closes when credits run out or the 91 days expire. **You will not be charged** unless you manually upgrade to a paid account. Credit card required for identity verification only (temporary $0–$1 hold, released within days).

---

## Registration (Step-by-Step)

1. Go to [console.cloud.google.com](https://console.cloud.google.com) or [cloud.google.com/free](https://cloud.google.com/free)
2. Sign in with any **Google account** (personal Gmail works fine)
3. Click **"Start free"** or **"Get started for free"**
4. Select your country and accept the Terms of Service
5. **Credit card verification** — enter a valid credit or debit card. Google places a temporary authorization hold of **$0–$1 USD** (not a charge; released within a few days to a month depending on your bank). Prepaid cards may not work. Make sure your card allows **automatic payments** — some banks (especially in India due to RBI regulations) block recurring authorizations and your signup will fail
6. Done — your Free Trial billing account is created with $300 in Welcome credits

**Important:**
- One free trial per person/organization — if you've previously been a paying customer of Google Cloud, Google Maps Platform, or Firebase, you don't qualify
- Multiple users within the same organization can each sign up for their own free trial
- Personal credit cards are fine; corporate cards are not required
- The 91-day clock starts at signup, not at first usage

---

## Free Trial Restrictions

During the free trial, you **cannot**:
- Add **GPUs** to VM instances (GPUs unlock only after upgrading to paid)
- Use **Google Cloud Marketplace** products
- Request a **quota increase**
- Access **third-party generative AI models** (model-as-a-service partners)
- Create **Windows Server** VM instances
- Use more than **8 concurrent Compute Engine CPU cores**
- Create Google Cloud VMware Engine resources

These restrictions lift immediately when you upgrade to a paid account. You keep any remaining credits after upgrading.

---

## Available AI Services

### Gemini API (via Google AI Studio — separate free tier, no credit card needed)

The Gemini API has its own **permanent free tier** through [Google AI Studio](https://aistudio.google.com) that is independent of the $300 trial. Free tier models include:

| Model | Free Tier | Notes |
|---|---|---|
| **Gemini 3 Flash Preview** | Free of charge | Latest model |
| **Gemini 2.5 Flash** | Free of charge | Production-ready |
| **Gemini 2.5 Flash-Lite** | Free of charge | Lightweight |
| **Gemini 2.5 Flash Native Audio** | Free of charge | Audio processing |
| **Gemini 2.5 Flash TTS** | Free of charge | Text-to-speech |
| **Gemini 2.0 Flash** | Free of charge | Retiring March 2026 |
| **Gemma 3 / Gemma 3n** | Free of charge | Open models |
| **Gemini Embedding** | Free of charge | Text embeddings |

Free tier rate limits are approximately **5–15 RPM** (requests/minute) and **up to 1,000 RPD** (requests/day) depending on model. Limits apply per Google Cloud project, not per API key. The free tier **allows commercial use**.

### Vertex AI (uses $300 trial credits)

- **Vertex AI Studio** — prompt design, model tuning, and testing with Gemini models
- **Vertex AI Express Mode** — simplified signup for @gmail.com accounts, separate 90-day free tier with limited quotas (up to 10 agent engines). This is separate from the $300 trial
- **Vertex AI Search** — 10,000 queries/month free
- **Vertex AI Pipelines, TensorBoard** — free while in preview
- Services in preview have 100% discounted pricing (free), but may transition to paid without much notice

**Note:** GPUs on Vertex AI are **not available** during the free trial. You must upgrade to a paid account first.

---

## AI-Related Always-Free Services

These are **perpetual** monthly limits that don't consume your $300 credit and persist even after the trial ends:

| Service | Monthly Free Limit |
|---|---|
| **Cloud Natural Language API** | 5,000 units |
| **Cloud Vision API** | 1,000 units |
| **Speech-to-Text** | 60 minutes |
| **Video Intelligence API** | 1,000 units |
| **Cloud Translation** | 500,000 characters |
| **Vertex AI Vizier** | First 100 trials (RANDOM_SEARCH/GRID_SEARCH) |

---

## Non-AI Always-Free Services (Highlights)

These also persist after the trial and don't consume credits:

| Service | Monthly Free Limit |
|---|---|
| **Compute Engine** | 1 e2-micro VM (us-east1, us-west1, us-central1 only), 30 GB disk |
| **Cloud Storage** | 5 GB regional (US regions), 100 GB egress |
| **BigQuery** | 1 TB queries, 10 GB storage |
| **Cloud Run** | 2M requests, 360K GB-seconds |
| **Cloud Functions** | 2M invocations, 400K GB-seconds |
| **Firestore** | 1 GB storage, 50K reads/day |
| **App Engine** | 28 instance-hours/day |
| **Pub/Sub** | 10 GB messages |
| **Cloud Shell** | 5 GB persistent storage |
| **Cloud Build** | 120 build-minutes/day |

---

## What Happens When the Trial Ends

1. **If you don't upgrade:** Your billing account auto-closes. All running workloads are **stopped** (not deleted immediately). You have a **30-day grace period** to upgrade and recover your resources. After 30 days, workloads are **permanently deleted**.
2. **If you upgrade to paid:** You keep any remaining credits until the 91-day expiration. You continue to get always-free tier usage. All trial restrictions (GPUs, Marketplace, etc.) are lifted. You only pay for usage beyond free tier limits.
3. **Always-free services continue** regardless — they don't require a paid account.

---

## Additional Tips

- **Gemini API via AI Studio vs. Vertex AI** — these are separate platforms with different billing, rate limits, and model access. AI Studio is free and doesn't require GCP billing. Vertex AI requires a GCP project with billing enabled (even if using free credits)
- **Google for Startups Cloud Program** — if you're building a startup, you can get up to $200K in credits (up to $350K for AI startups) through the [startup program](https://cloud.google.com/startup)
- **Google Cloud Teaching Credits** — faculty can apply for $50–$100/semester per student for coursework
- **Google Cloud Research Credits** — up to $5,000 for faculty/postdocs, $1,000 for PhD students
- **Region matters** — always-free compute and storage are limited to US regions (us-east1, us-west1, us-central1)
- **Set budget alerts** — if you upgrade to paid, set up budget alerts in the Billing console to avoid surprise charges. Google Cloud has no spending caps by default
- **Gemini 2.0 Flash retirement** — models are being retired March 3, 2026. Migrate to Gemini 2.5 Flash or newer

---

**Sources:**
- [Google Cloud Free Program](https://cloud.google.com/free)
- [Google Cloud Free Features Documentation](https://docs.cloud.google.com/free/docs/free-cloud-features)
- [Google Cloud Free Trial FAQs](https://cloud.google.com/signup-faqs)
- [Google Cloud Free Trial Terms](https://cloud.google.com/terms/free-trial)
- [Billing During Free Trial](https://support.google.com/cloud/answer/7006543)
- [Gemini API Pricing](https://ai.google.dev/gemini-api/docs/pricing)
- [Gemini API Rate Limits](https://ai.google.dev/gemini-api/docs/rate-limits)
- [Vertex AI Express Mode](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/start/express-mode/overview)
- [Vertex AI Pricing](https://cloud.google.com/vertex-ai/pricing)

## Submission 30: Google Cloud Always Free Tier

- Category: Cloud Providers
- Source Link: https://cloud.google.com/free
- Amount: Usage-limited free tier (no dollar value; monthly quotas per service). Key highlights: 1 e2-micro VM, 5 GB Cloud Storage, 1 TB BigQuery queries, 2M Cloud Run requests, and Gemini API access with 100-1,000 requests/day depending on model.
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Google Cloud provides **perpetual free usage** of 25+ products with monthly limits that reset every billing cycle. Unlike the $300 trial (which expires after 91 days), the Always Free tier **never expires** and continues even after upgrading to a paid account. These limits are per billing account, not per project.

**Credit card required?** Yes, to create a GCP billing account. However, the Gemini API free tier via Google AI Studio does **not** require a credit card.

---

## Registration

1. Go to [cloud.google.com/free](https://cloud.google.com/free)
2. Sign in with a Google account
3. Click **"Get started for free"** and provide a credit/debit card for verification
4. A temporary $0-$1 authorization hold is placed (released within days)
5. Your billing account is created with access to Always Free products immediately

Alternatively, for Gemini API only: go to [aistudio.google.com](https://aistudio.google.com), sign in with Google, and get an API key — no billing account or credit card needed.

---

## AI-Related Free Services

### Gemini API (via Google AI Studio — no credit card needed)

The Gemini API has a **permanent free tier** through Google AI Studio, separate from GCP billing. All free-tier models support a **1 million token context window** and **allow commercial use**.

| Model | RPM | TPM | RPD | Best For |
|---|---|---|---|---|
| **Gemini 2.5 Pro** | 5 | 250,000 | 100 | Complex reasoning |
| **Gemini 2.5 Flash** | 10 | 250,000 | 250 | General-purpose |
| **Gemini 2.5 Flash-Lite** | 15 | 250,000 | 1,000 | High-throughput simple tasks |
| **Gemini 2.0 Flash** | 5 | 250,000 | — | Legacy (retiring March 2026) |
| **Gemini 3 Flash Preview** | Free | — | — | Latest preview |
| **Gemini 2.5 Flash TTS** | Free | — | — | Text-to-speech |
| **Gemma 3 / Gemma 3n** | Free | — | — | Open-weight models |
| **Gemini Embedding** | Free | — | — | Text embeddings |

*RPM = Requests/minute, TPM = Tokens/minute, RPD = Requests/day. Limits are per GCP project, not per API key. Daily quotas reset at midnight Pacific Time.*

### GCP AI/ML APIs (require billing account)

| Service | Monthly Free Limit |
|---|---|
| **Cloud Natural Language API** | 5,000 units |
| **Cloud Vision API** | 1,000 units |
| **Speech-to-Text** | 60 minutes |
| **Video Intelligence API** | 1,000 units |

---

## Non-AI Free Services

### Compute

| Service | Free Limit | Notes |
|---|---|---|
| **Compute Engine** | 1 e2-micro VM/month, 30 GB standard disk, 1 GB egress | **US regions only** (us-east1, us-west1, us-central1) |
| **Cloud Run** | 2M requests, 180,000 vCPU-seconds (50 hrs), 360,000 GiB-seconds (100 hrs), 1 GB egress/month | — |
| **Cloud Run Functions** | 2M invocations, 400,000 GB-seconds, 200,000 GHz-seconds, 5 GB egress/month | — |
| **App Engine** | 28 instance-hours/day (F1), 9 instance-hours/day (B1), 1 GB egress/day | — |
| **Google Kubernetes Engine** | 1 Autopilot or Zonal cluster (no cluster management fee) | Node resources billed separately |
| **Agent Engine** | 180,000 vCPU-seconds, 360,000 GiB-seconds/month | — |

### Storage & Databases

| Service | Free Limit | Notes |
|---|---|---|
| **Cloud Storage** | 5 GB standard storage, 5,000 Class A ops, 50,000 Class B ops, 100 GB egress/month | **US regions only** |
| **BigQuery** | 1 TB queries, 10 GB storage/month | BigQuery Sandbox available without credit card |
| **Firestore** | 1 GB storage, 50,000 reads, 20,000 writes, 20,000 deletes/day, 10 GB egress/month | — |
| **Artifact Registry** | 0.5 GB storage/month | — |

### Networking & Messaging

| Service | Free Limit |
|---|---|
| **Pub/Sub** | 10 GB messages/month |

### DevOps & Security

| Service | Free Limit |
|---|---|
| **Cloud Build** | 2,500 build-minutes/month |
| **Cloud Source Repositories** | 5 users, 50 GB storage, 50 GB egress/month |
| **Cloud Shell** | Free access with 5 GB persistent disk |
| **Secret Manager** | 6 active versions, 10,000 access ops, 3 rotation notifications/month |
| **Cloud KMS** | 100 active key versions, 10,000 operations/month |
| **reCAPTCHA** | 10,000 assessments/month |
| **Web Risk** | 100,000 URI searches/month |

### Other

| Service | Free Limit |
|---|---|
| **Cloud Observability (Logging)** | 50 GB logs/project/month, 1M API read calls |
| **Workflows** | 5,000 internal steps, 2,000 HTTP calls/month |
| **Application Integration** | 400 executions, 20 GB data, 2 connection nodes/month |
| **Cloud Deploy** | 1 active delivery pipeline/account |
| **Workload Manager** | 5,000 resource evaluations/month |

---

## Region Restrictions

- **Compute Engine e2-micro**: Free only in **us-west1** (Oregon), **us-central1** (Iowa), and **us-east1** (South Carolina). Deploying in any other region incurs charges.
- **Cloud Storage 5 GB**: Free only in **US regions**.
- **All other services**: Generally available globally, but always verify in the official documentation since region restrictions can change.

---

## What Happens If You Exceed Limits

- **With a paid billing account:** You are automatically billed at standard rates for any usage beyond the free tier. There is no hard cap — overages simply become billable usage.
- **With a Free Trial account:** Overages consume your $300 credit. You cannot be charged beyond the credit amount until you manually upgrade.
- **BigQuery Sandbox (no billing account):** Queries simply fail when the free limit is reached — no billing possible.
- **Gemini API free tier:** Returns HTTP 429 (rate limit exceeded) errors. No billing occurs.

---

## Additional Tips

- **Set budget alerts immediately** if you upgrade to a paid account. Google Cloud has no default spending caps. Go to Billing > Budgets & Alerts and create a $0 budget to get notified of any charges.
- **Gemini API quotas are per project** — creating multiple API keys in the same project does not multiply your limits. Create separate GCP projects if you need more quota.
- **BigQuery Sandbox** is the easiest entry point — no credit card needed, just enable the sandbox from the BigQuery console.
- **December 2025 quota reduction** — Google reduced Gemini free tier quotas by 50-80% in December 2025. If your app relied on previous limits, you may now hit rate limits more frequently.
- **Route Gemini requests by complexity** — use Flash-Lite (15 RPM, 1,000 RPD) for simple tasks, Flash (10 RPM) for moderate work, and reserve Pro (5 RPM, 100 RPD) for complex reasoning.
- **Google reserves the right to change** Always Free limits with 30 days advance notice. The tier has no contractual end date but quotas can shift.
- **Google for Startups Cloud Program** offers up to $200K in credits ($350K for AI startups) — apply at [cloud.google.com/startup](https://cloud.google.com/startup).
- **Academic credits** are available: $50-$100/semester per student for faculty, up to $5,000 for research.
- **Gemini 2.0 Flash** is being retired on March 3, 2026. Migrate to Gemini 2.5 Flash or newer.

---

**Sources:**
- [Google Cloud Free Program](https://cloud.google.com/free)
- [Google Cloud Free Features Documentation](https://docs.cloud.google.com/free/docs/free-cloud-features)
- [Google Cloud Free Trial FAQs](https://cloud.google.com/signup-faqs)
- [Gemini API Pricing](https://ai.google.dev/gemini-api/docs/pricing)
- [Gemini API Free Tier Rate Limits Guide](https://www.aifreeapi.com/en/posts/gemini-api-free-tier-rate-limits)
- [Google Cloud Billing — Free Trial](https://support.google.com/cloud/answer/7006543)

## Submission 29: GitHub Student Developer Pack - $500+ Value

- Category: Education & Student
- Source Link: https://education.github.com/pack
- Amount: $500+ combined value across all partner offers
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

The GitHub Student Developer Pack bundles **80+ professional developer tools and cloud credits** into a single free package for verified students. Highlights include **GitHub Copilot Pro** (normally $10/month), **$200 in DigitalOcean credits**, **$100 in Azure credits**, **JetBrains IDE subscriptions**, a **free domain name**, and partner offers from MongoDB, Heroku, Datadog, New Relic, 1Password, and dozens more. It is one of the highest-value free programs available to students — no credit card required.

---

## Registration (Step-by-Step)

1. Go to [education.github.com/pack](https://education.github.com/pack)
2. Click **"Get your Pack"** (or **"Get benefits"**)
3. Sign in with your **GitHub account** (create one first if you do not have one)
4. Select **"Student"** and choose your school from the dropdown
5. **Verify your student status** — use one of the following:
   - **School-issued email** (e.g. `@university.edu`) — fastest path, often auto-approved
   - **Upload proof of enrollment** — student ID card, transcript, enrollment letter, or class schedule. The document must clearly show your **name**, **school name**, and a **current date**
6. Enter your **expected graduation year** and a brief description of how you plan to use the tools
7. Click **"Submit application"**
8. Wait for approval — typically **a few days**, but can take **1-2 weeks** during peak periods (semester starts)
9. Once approved, you receive an email. Go to the Pack page and **activate each partner offer individually**

**Important:**
- **No credit card required** for the Pack itself (individual partners may have their own requirements)
- **Disable your VPN** during the application — GitHub uses location data for verification
- **Allow browser location access** when prompted
- Submit **clear, unedited photos** of documents — cropped or edited images get flagged as manipulated
- If rejected, you can resubmit with better documentation. Common rejection reasons: missing date on documents, unclear school name, or location mismatch

---

## Student Verification Details

GitHub accepts several forms of proof, in order of preference:

| Document Type | Requirements | Approval Speed |
|---|---|---|
| **School email** (.edu or institutional) | Must be your primary or added GitHub email | Fastest (often automatic) |
| **Student ID card** | Photo, full name, school name, current date visible | 1-7 days |
| **Enrollment letter** | Official letterhead, current semester/term | 1-7 days |
| **Transcript** | Name, school, recent dates | 1-7 days |
| **Class schedule** | Current term, your name, school name | 1-7 days |

If your school email is not recognized as academic, you will be prompted to upload additional documentation. Bootcamp and online program students can qualify if they can provide official enrollment proof.

---

## GitHub Copilot Pro (AI Coding Assistant)

Verified students get **GitHub Copilot Pro for free** (normally $10/month). This is the full Pro plan, not a limited version.

### What Copilot Pro includes:
- **Unlimited code completions** in your editor (VS Code, JetBrains, Neovim, etc.)
- **300 premium requests/month** for Copilot Chat, agent mode, and code reviews
- **Multi-model access** — GPT-4.1, GPT-5 mini, Claude Sonnet, Gemini Flash, and more
- **Copilot coding agent** — autonomous task execution from GitHub Issues
- **Copilot Chat** in IDE and on github.com
- **Copilot CLI** — terminal command suggestions

### Models available in Copilot Pro:

| Model | Premium Request Cost | Notes |
|---|---|---|
| **GPT-4.1** | Included (no cost) | Default for completions |
| **GPT-5 mini** | Included (no cost) | Fast, capable |
| **GPT-4o** | 1 request | Multimodal |
| **Claude Sonnet 4.5** | 1 request | Strong for code |
| **Claude Haiku 4.5** | 0.25 requests | Fast, lightweight |
| **Gemini 2.5 Flash** | 0.25 requests | Very fast |
| **GPT-5** | 5 requests | Deep reasoning |
| **GPT-5.2** | 5 requests | Latest flagship |
| **Claude Opus 4.5** | 10 requests | Most capable, expensive |

### What students do NOT get:
- **Copilot Pro+** ($39/month) features: 1,500 premium requests, access to all models including GPT-5.2-Codex and GPT-5.1-Codex
- The student benefit **does not stack** as a $10 credit toward Pro+

---

## Key Partner Offers

### Cloud & Infrastructure

| Partner | Offer | Duration |
|---|---|---|
| **DigitalOcean** | $200 in platform credit | 1 year |
| **Microsoft Azure** (18+) | $100 Azure credit + 25 free services | 12 months |
| **Microsoft Azure** (13-17) | Free App Services, Functions, MySQL, DevOps | While student |
| **Heroku** | $13/month credit | 24 months |
| **New Relic** | Free account ($300/month value) | While student |
| **Datadog** | Pro Account with 10 servers | 2 years |

### Developer Tools & IDEs

| Partner | Offer | Duration |
|---|---|---|
| **JetBrains** | All IDEs (IntelliJ, PyCharm, WebStorm, etc.) | Annual renewal |
| **GitHub Pro** | Upgraded account features | While student |
| **GitHub Codespaces** | Free Pro-level access | While student |
| **GitKraken** | 6 months free, then up to 80% off | 6 months |
| **GitLens** | 6 months free, then up to 80% off | 6 months |
| **Visual Studio Dev Essentials** | VS Community + Pluralsight + Azure credit | Ongoing |
| **Bootstrap Studio** | Free license | While student |
| **Tower** | Free Tower Pro license | While student |

### Domains & Hosting

| Partner | Offer | Duration |
|---|---|---|
| **Namecheap** | Free .me domain + SSL certificate | 1 year |
| **Name.com** | Free domain (25+ extensions: .live, .app, .dev) | 1 year |
| **.TECH** | Free .tech domain | 1 year |
| **GitHub Pages** | Free static site hosting | Ongoing |

### Databases & Backend

| Partner | Offer | Duration |
|---|---|---|
| **MongoDB** | $50 Atlas credits + free certification | One-time |
| **Appwrite** | Free Education plan | While student |
| **Doppler** | Free Team subscription (secrets management) | While student |

### Learning & Education

| Partner | Offer | Duration |
|---|---|---|
| **FrontendMasters** | Full access to all courses | 6 months |
| **Educative** | Free access + 30% discount | 6 months |
| **DataCamp** | Full subscription access | 3 months |
| **Codedex** | Club premium membership | 6 months |
| **Scrimba** | Full access to 40+ courses | 1 month |
| **GoRails** | All videos and lessons | 12 months |
| **AlgoExpert** | 20 interview questions + 10% off | Ongoing |
| **InterviewCake** | Full interview prep course | 1 week |

### Design & Productivity

| Partner | Offer | Duration |
|---|---|---|
| **Notion** | Education plan with AI responses | While student |
| **Canva** | Free Canva Pro features | While student |
| **1Password** | Free account + Developer Tools | 1 year |
| **Icons8** | Icons, photos, illustrations, music | 3 months |
| **IconScout** | 60 premium icons/month | 1 year |
| **Visme** | Starter plan access | 3 months |

### DevOps & Testing

| Partner | Offer | Duration |
|---|---|---|
| **Sentry** | 50K errors, 100K transactions, 1GB attachments | 1 year |
| **BrowserStack** | Automate Mobile Plan | 1 year |
| **LambdaTest** | Live Plan | 1 year |
| **Codecov** | Free on public and private repos | While student |
| **Travis CI** | Free private builds | While student |
| **Testmail** | Essential plan | While student |
| **Blackfire** | Developer subscription | While student |

### Security

| Partner | Offer | Duration |
|---|---|---|
| **Astra Security** | Website firewall + malware scanner | 6 months |
| **Honeybadger** | Small account | 1 year |

---

## GitHub Certifications

Verified students receive **1 free voucher** for a GitHub Certification exam:
- **GitHub Foundations** certification, or
- **GitHub Copilot** certification

Exams are administered through **Microsoft Learn / Pearson VUE**. Current vouchers expire **June 30, 2026**.

---

## Duration, Renewal & Expiration

- Benefits are valid for **2 years** from approval date
- You can **renew** by resubmitting updated proof of enrollment once benefits expire
- Renewal is treated like a new application — you need current documentation
- Some partner offers **cannot be renewed** (e.g., Namecheap domain is for first-time users only)
- After **graduation**, you lose student status and associated benefits, but keep your GitHub account, repos, and contribution history
- If a partner leaves the Pack while you have an active offer, you can typically complete the remaining term

---

## Additional Tips

- **Activate offers promptly** — some partner offers have their own expiration windows that start when you claim them, not when you get the Pack
- **Some partners have age restrictions** — Azure credits require age 18+; the 13-17 tier has different (more limited) services
- **Regional restrictions apply** — some partners may not serve all countries regardless of your student verification status
- **Two-factor authentication (2FA)** may be required on your GitHub account before you can apply
- **GitHub Campus Experts** — if you are a community leader at your school, you can apply to the Campus Experts program for additional mentorship and resources
- **Microsoft for Startups Founders Hub** — if you are building a startup alongside your studies, this separate program offers up to $5,000 in Azure credits
- **Keep your student email active** — some benefits verify status through your email periodically
- **Offers change over time** — partners join and leave the Pack. Check the Pack page regularly for new additions

---

**Sources:**
- [GitHub Student Developer Pack](https://education.github.com/pack)
- [GitHub Education - Students](https://github.com/education/students)
- [Student Developer Pack Application FAQs](https://github.com/orgs/community/discussions/111352)
- [Student Developer Pack Partners FAQ Repo](https://github.com/github-education-resources/Student-Developer-Pack-Current-Partners-FAQ)
- [GitHub Copilot Plans & Pricing](https://github.com/features/copilot/plans)
- [Supported AI Models in GitHub Copilot](https://docs.github.com/en/copilot/reference/ai-models/supported-models)
- [DigitalOcean GitHub Students](https://www.digitalocean.com/github-students)
- [Heroku GitHub Students Offer](https://www.heroku.com/github-students/)
- [How to Get GitHub Student Developer Pack (Apidog)](https://apidog.com/blog/free-github-student-developer-pack/)

## Submission 28: GitHub Models - Free Prototyping Tier

- Category: AI API Free Tiers
- Source Link: https://github.com/marketplace/models
- Amount: Rate-limited free access (e.g., 50 requests/day for GPT-4o High tier; 150 requests/day for Low tier models)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

GitHub Models gives every GitHub user **free, rate-limited access to 40+ AI models** from OpenAI, Meta, DeepSeek, Mistral, Microsoft, Cohere, xAI, and AI21 Labs. No credit card, no Azure account, no separate API key signup — just your GitHub login. The free tier is designed for **prototyping and experimentation**, with strict per-model rate limits (e.g., 50 requests/day for GPT-4o). When you are ready for production, you can opt into pay-as-you-go billing or bring your own API key (BYOK).

---

## Registration (Step-by-Step)

1. Go to [github.com/marketplace/models](https://github.com/marketplace/models)
2. **Sign in** with any GitHub account (free accounts work)
3. **Playground** — click any model to open an interactive chat playground. Works immediately, no setup required
4. **API access** — go to **Settings → Developer settings → Personal access tokens → Fine-grained tokens**
5. Generate a new token with the **`models:read`** scope (this scope is required since May 2025)
6. Use the token with the API endpoint: `https://models.github.ai/inference/chat/completions`
7. Set the `Authorization: Bearer <YOUR_TOKEN>` header in your requests

**Important:**
- No credit card or payment method required for the free tier
- One GitHub account = one set of free rate limits
- The playground works in-browser with zero configuration
- In GitHub Actions workflows, add `models: read` to the permissions block and use the built-in `GITHUB_TOKEN`

---

## Available AI Models

GitHub Models provides access to **40+ models** across multiple providers. The catalog is continuously updated. Key models include:

### OpenAI
| Model | Type |
|---|---|
| **GPT-5.2** | Latest flagship |
| **GPT-5.2-Codex** | Latest code-specialized |
| **GPT-5** / GPT-5.1 | Flagship models |
| **GPT-5 mini** | Compact flagship |
| **GPT-4.1** / 4.1 mini / 4.1 nano | Production models (1M token context) |
| **GPT-4o** / 4o mini | Multimodal models |
| **o4-mini** | Reasoning (improved o3-mini) |
| **o3** / o3-mini | Reasoning models |
| **o1** | Advanced reasoning |
| **Text Embedding 3** (large/small) | Embedding models |

### Meta (Llama)
| Model | Type |
|---|---|
| **Llama 4 Maverick 17B 128E** | Latest MoE model |
| **Llama 4 Scout 17B 16E** | Multi-doc summarization |
| **Llama 3.3 70B Instruct** | General purpose |
| **Llama 3.2 11B/90B Vision** | Multimodal |
| **Llama 3.1 405B / 8B** | Large and small variants |

### DeepSeek
| Model | Type |
|---|---|
| **DeepSeek-R1** / R1-0528 | Reasoning model |
| **DeepSeek-V3-0324** | General purpose |
| **MAI-DS-R1** | Microsoft-tuned DeepSeek-R1 |

### Mistral AI
| Model | Type |
|---|---|
| **Mistral Medium 3** | Mid-size model |
| **Mistral Small 3.1** | Multimodal, 128K context |
| **Ministral 3B** | Ultra-compact |
| **Codestral 25.01** | Code-specialized |

### Microsoft (Phi)
| Model | Type |
|---|---|
| **Phi-4 14B** | Low-latency SLM |
| **Phi-3.5 / Phi-4 mini** | 3.8B parameter SLM |

### Anthropic
| Model | Type |
|---|---|
| **Claude Opus 4.5** | Most capable |
| **Claude Sonnet 4.5** | Balanced performance |
| **Claude Haiku 4.5** | Fast, lightweight |

### Other Providers
| Model | Provider |
|---|---|
| **Grok 3** / Grok 3 Mini | xAI |
| **Cohere Command A** (398B MoE) | Cohere |
| **Cohere Command R / R+** | Cohere |
| **AI21 Jamba 1.5 Large** | AI21 Labs |

Browse the full, up-to-date catalog at [github.com/marketplace/models](https://github.com/marketplace/models) or use the REST API: `GET /catalog/models`.

---

## Free Tier Rate Limits

Each model is classified into a **rate limit tier** (Low, High, or Embedding). You can find a model's tier on its marketplace page.

| Tier | Requests/min | Requests/day | Tokens/request (input) | Tokens/request (output) | Concurrent requests |
|---|---|---|---|---|---|
| **High** (e.g., GPT-4o, GPT-5, o3) | 10 | 50 | 8,000 | 4,000 | 2 |
| **Low** (e.g., Phi-4, Llama 3.1 8B, Ministral 3B) | 15 | 150 | 8,000 | 4,000 | 5 |
| **Embedding** (e.g., Text Embedding 3) | 15 | 150 | 64,000 | — | 5 |

**Key details:**
- Limits apply per GitHub account, not per token or project
- Limits may vary based on your **GitHub Copilot plan** (Free, Pro, Pro+, Business, Enterprise)
- Limits are subject to change without notice
- When you hit the limit, requests return HTTP 429 until the window resets
- Token limits are **per request**, not per day — you cannot batch beyond the per-request cap

---

## Copilot Plan Interaction

GitHub Models rate limits are connected to your GitHub Copilot plan. While all GitHub accounts get free access, your plan may affect the exact limits:

| Copilot Plan | Price | Notes |
|---|---|---|
| **Free** | $0/month | Base rate limits for GitHub Models |
| **Pro** | $10/month | Higher limits, 300 premium requests/month for Copilot features |
| **Pro+** | $39/month | Highest individual limits, 1,500 premium requests/month |
| **Business** | $19/user/month | Organization-level, 300 premium requests/month |
| **Enterprise** | $39/user/month | Full enterprise features, 1,000 premium requests/month |

**Note:** Copilot premium requests and GitHub Models rate limits are related but separate concepts. The Copilot premium request allowance covers Copilot Chat and coding agent usage, while GitHub Models rate limits govern direct API/playground inference.

---

## Scaling Beyond Free Limits

When you outgrow the free tier, there are two paths:

### 1. Pay-as-you-go (GitHub billing)
- Opt in through your GitHub account settings
- Unified pricing: **$0.00001 USD per token unit**
- Token units = (input tokens x input multiplier) + (output tokens x output multiplier)
- Multipliers vary by model (e.g., GPT-4o: 0.25x input, 1x output)
- Example: 1M input + 1M output tokens on GPT-4o = ~$12.50
- Billing is through your GitHub account, same as Copilot

### 2. Bring Your Own Key (BYOK)
- Connect your own OpenAI or Azure AI API key
- Billing goes directly to your provider account
- Access custom models and higher rate limits set by the provider
- Rate limits are governed by your provider subscription, not GitHub

---

## Developer Tooling

GitHub Models is more than just an API — it includes developer workflow features:

- **Playground** — interactive chat UI for testing any model with adjustable temperature, token limits, and system prompts
- **Prompt files** — save `.prompt.yml` files in your repo for version-controlled prompts with variables
- **Evaluations** — add `testData` and `evaluators` to prompt files for side-by-side model comparison
- **GitHub Actions integration** — call models from CI/CD workflows using the built-in `GITHUB_TOKEN`
- **REST API** — `GET /catalog/models` to list models, `POST /inference/chat/completions` for inference
- **Azure AI Inference SDK** — compatible with the same SDK used for Azure OpenAI

---

## Additional Tips

- **No separate Azure account needed** — GitHub Models is powered by Azure AI under the hood, but you never need to create Azure resources for the free tier
- **Same API format as OpenAI** — the inference endpoint uses the OpenAI-compatible chat completions format, so existing code often works with minimal changes
- **Model switching is easy** — just change the `model` parameter in your API call; the endpoint and auth stay the same
- **Students and OSS maintainers** — qualifying users get free GitHub Copilot Pro, which may include enhanced GitHub Models limits
- **GitHub Actions caveat** — model calls from Actions count against your account's rate limits just like direct API calls
- **Data privacy** — prompts and outputs are processed within GitHub/Azure infrastructure; check GitHub's data policies for specifics on retention
- **Limits are strict** — the 8K input / 4K output token cap per request means you cannot use the free tier for long-context tasks (e.g., summarizing large documents)

---

**Sources:**
- [GitHub Models Marketplace](https://github.com/marketplace/models)
- [GitHub Models Features](https://github.com/features/models)
- [Prototyping with AI Models - GitHub Docs](https://docs.github.com/en/github-models/prototyping-with-ai-models)
- [GitHub Models Quickstart](https://docs.github.com/en/github-models/quickstart)
- [GitHub Models Billing](https://docs.github.com/billing/managing-billing-for-your-products/about-billing-for-github-models)
- [GitHub Models API - models:read Required](https://github.blog/changelog/2025-05-15-modelsread-now-required-for-github-models-access/)
- [GitHub Models Pay-as-you-go](https://github.blog/changelog/2025-06-24-github-models-now-supports-moving-beyond-free-limits/)
- [GitHub Copilot Plans & Pricing](https://github.com/features/copilot/plans)
- [REST API for Models Inference](https://docs.github.com/en/rest/models/inference)

## Submission 27: Fireworks AI - $1 Free Credits for New Users

- Category: AI API Free Tiers
- Source Link: https://fireworks.ai/
- Amount: $1 in free credits
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Fireworks AI gives every new user **$1 in free credits** to explore their inference platform — no credit card required. The platform specializes in fast, optimized inference for open-source models including DeepSeek, Llama, Qwen, Mixtral, GLM, and many others, with serverless pricing starting at **$0.10 per 1M tokens** for small models. After credits run out, you add a payment method and pay as you go. The infrastructure is **HIPAA and SOC 2 Type II compliant**, with zero data retention by default.

---

## Registration (Step-by-Step)

1. Go to [fireworks.ai/signup](https://fireworks.ai/signup)
2. Sign in with your **Google account** (default SSO) or create an account with email
3. Your account is created with **$1 in free credits** automatically applied
4. Go to the [API Keys](https://app.fireworks.ai/api-keys) section in the dashboard to generate your API key
5. Start making API calls using the OpenAI-compatible endpoint: `https://api.fireworks.ai/inference/v1`

**Important:**
- **No credit card required** to sign up and start using the free credits
- Without a payment method on file, you are limited to **10 requests per minute (RPM)**
- Adding a payment method unlocks up to **600 RPM** (with a 6,000 RPM spike arrest ceiling)
- If you do not add a payment method, your account is **suspended** once the $1 credit is exhausted — no surprise charges

---

## Available Models

Fireworks AI hosts hundreds of open-source models. Key categories and examples:

### Text / Chat Models

| Model | Input Price (per 1M tokens) | Output Price (per 1M tokens) |
|---|---|---|
| **DeepSeek R1** | $1.35 | $5.40 |
| **DeepSeek V3** | $0.56 | $1.68 |
| **Qwen3 235B** | $0.22 | $0.88 |
| **Kimi K2.5** | $0.60 | $3.00 |
| **GLM-4.7** | $0.60 | $2.20 |
| **GLM-4.6** | $0.55 | $2.19 |
| **Llama 4 / 3.3 / 3.1** | $0.10 – $0.90 | varies by size |
| **Mixtral 8x7B** (MoE) | $0.50 | $0.50 |
| Models <4B params | $0.10 | $0.10 |
| Models 4B–16B params | $0.20 | $0.20 |
| Models 16B+ params | $0.90 | $0.90 |

### Vision / Multimodal Models

- Qwen3 VL series, Llama 3.2 Vision, GLM-4.5V, Molmo2, InternVL3

### Image Generation

| Model | Price |
|---|---|
| **FLUX Kontext Pro** | $0.04/image |
| **FLUX Kontext Max** | $0.08/image |
| **FLUX.1 [schnell]** | ~$0.0014/image |
| **FLUX.1 [dev]** | ~$0.014/image |
| **Stable Diffusion XL** | ~$0.0039/image |

### Audio (Speech-to-Text)

| Model | Price |
|---|---|
| **Whisper V3 Large** | $0.0015/minute |
| **Whisper V3 Turbo** | $0.0009/minute |

Diarization adds a 40% surcharge. Batch API reduces audio pricing by 40%.

### Embeddings & Reranking

| Model Size | Price (per 1M tokens) |
|---|---|
| Up to 150M params | $0.008 |
| 150M–350M params | $0.016 |
| Qwen3 8B Embedding | $0.10 |

Reranking models (Qwen3 Reranker) are also available.

---

## How Far Does $1 Go?

With $1 in free credits, here is a rough estimate of what you can do:

| Use Case | Approximate Usage |
|---|---|
| Qwen3 235B input tokens | ~4.5M tokens (~3,000 pages of text) |
| Small model (<4B) tokens | ~10M tokens |
| DeepSeek V3 input tokens | ~1.8M tokens |
| FLUX.1 [schnell] images | ~714 images |
| Whisper V3 Turbo transcription | ~18.5 hours of audio |
| Embeddings (small model) | ~125M tokens |

This is enough for meaningful prototyping and experimentation, but not for sustained production use.

---

## Spend Limits & Tier System

Fireworks uses a tier system that controls your maximum monthly spend:

| Tier | Requirement | Max Monthly Budget |
|---|---|---|
| **Free** (no payment method) | Signup only | $1 (free credits only) |
| **Tier 1** | Valid payment method | $50/month |
| **Tier 2** | $50+ in spend or credits | $500/month |
| **Tier 3** | $500+ in spend or credits | $5,000/month |
| **Tier 4** | $5,000+ in spend or credits | $50,000/month |
| **Unlimited** | Contact sales | Unlimited |

- Adding prepaid credits counts toward tier advancement (e.g., buying $100 in credits moves you from Tier 1 to Tier 2)
- When you hit your spend limit, **all API requests pause** — no overage charges
- Tier upgrades activate within minutes

---

## Rate Limits

| Account Type | Rate Limit |
|---|---|
| No payment method (free credits only) | **10 RPM** |
| With payment method (all tiers) | **Up to 600 RPM** (6,000 RPM spike ceiling) |
| Audio (prod) | 200 RPM |
| Audio (turbo) | 400 RPM |
| On-demand deployments | No hard RPM limits (GPU quota based) |

The 10 RPM limit without a payment method is restrictive for any real testing. Consider adding a credit card to unlock 600 RPM even if you only plan to use the free credits.

---

## API Compatibility

Fireworks AI uses an **OpenAI-compatible API format**, making it a drop-in replacement in most codebases. Change the base URL and API key, and existing OpenAI SDK code works:

```
Base URL: https://api.fireworks.ai/inference/v1
Model format: accounts/fireworks/models/<model-name>
```

Supported SDKs: OpenAI Python/Node, LangChain, LlamaIndex, Vercel AI SDK, LiteLLM, and others.

---

## Fine-Tuning

Fireworks supports supervised fine-tuning (SFT), preference tuning (DPO), and reinforcement fine-tuning:

| Model Size | SFT (per 1M tokens) | DPO (per 1M tokens) |
|---|---|---|
| Up to 16B | $0.50 | $1.00 |
| 16B–80B | $3.00 | $6.00 |
| 80B–300B | $6.00 | $12.00 |
| 300B+ | $10.00 | $20.00 |

Fine-tuned models serve at base model pricing. LoRA adapters can be deployed on shared or dedicated infrastructure.

---

## Security & Compliance

- **SOC 2 Type II** certified (third-party audited)
- **HIPAA** compliant — suitable for healthcare workloads with a BAA
- **Zero Data Retention** by default — prompts and outputs are not logged or stored unless you opt in
- **Encryption:** TLS 1.2+ in transit, AES-256 at rest
- **GDPR and CCPA** aligned
- **Bring Your Own Bucket** — use your own cloud storage for data governance
- Trust Center available at [trust.fireworks.ai](https://trust.fireworks.ai/)

---

## On-Demand GPU Deployments

For dedicated capacity beyond serverless, Fireworks offers per-second GPU billing:

| GPU Type | Cost per Hour |
|---|---|
| A100 80GB | $2.90 |
| H100 80GB | $4.00 |
| H200 141GB | $6.00 |
| B200 180GB | $9.00 |

These are billed per second with no startup charges. Useful for deploying custom models or getting guaranteed throughput.

---

## Additional Tips

- **$1 goes further on small models** — if you are just experimenting, use models under 4B parameters ($0.10/1M tokens) to maximize your free credits
- **Cached input tokens cost 50% less** — repeated prompts with shared prefixes save money
- **Batch inference is 50% off** — if latency is not critical, use the batch API to halve your costs
- **Account suspension is harmless** — if credits run out without a payment method, your account pauses with no charges. Add a card to resume
- **Discord community** — developer tier support is via [Discord](https://discord.gg/fireworks-ai), not email tickets
- **No startup or student programs** publicly listed — unlike major cloud providers, Fireworks does not have a separate startup credits program
- **Performance advantage** — Fireworks claims over 250 tokens/second throughput for DeepSeek models, which is significantly faster than many competitors

---

**Sources:**
- [Fireworks AI Pricing](https://fireworks.ai/pricing)
- [Fireworks AI Models](https://fireworks.ai/models)
- [Rate Limits & Quotas](https://docs.fireworks.ai/guides/quotas_usage/rate-limits)
- [Cost Structure](https://docs.fireworks.ai/faq/billing-pricing-usage/pricing/cost-structure)
- [Getting Started](https://docs.fireworks.ai/getting-started/introduction)
- [Data Security](https://docs.fireworks.ai/guides/security_compliance/data_security)
- [SOC 2 & HIPAA Compliance](https://fireworks.ai/blog/fireworks-ai-achieves-soc-2-type-ii-and-hipaa-compliance)
- [Fireworks AI Trust Center](https://trust.fireworks.ai/)

## Submission 26: Microsoft Azure Free Account - $200 Credits

- Category: General
- Source Link: https://azure.microsoft.com/en-us/pricing/purchase-options/azure-account
- Amount: $200 (30 days) + 12 months free services + 65+ always-free services
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

New Azure customers receive **$200 in credits for 30 days**, plus **12 months of free access** to 20+ popular services and **65+ always-free services**. Credits can be applied to any eligible Azure service including compute, databases, and AI services.

---

## Registration (Step-by-Step)

1. Go to [azure.microsoft.com/free](https://azure.microsoft.com/en-us/pricing/purchase-options/azure-account/)
2. Click **"Start free"**
3. Sign in with a **Microsoft account** or **GitHub account** (or create a new one)
4. **Phone verification** — enter your phone number, choose SMS or callback, enter the code (VoIP numbers not accepted)
5. **Credit card verification** — enter a valid credit/debit card (non-prepaid). You will NOT be charged; a temporary ~$1 hold appears and is removed within 3-5 days
6. **Accept** the Azure agreement and privacy statement
7. Done — you now have $200 credit + free tier access

**Important:**
- One free account per person (same email/phone/card triggers restrictions on re-registration)
- After 30 days or credit exhaustion, you must upgrade to pay-as-you-go to keep free-tier services
- No auto-shutoff after 12 months — services silently transition to paid rates, so set up budget alerts

---

## Azure OpenAI — Available AI Models

### No registration required (all Azure customers):
- **GPT-4o**, GPT-4o mini
- **GPT-4.1**, GPT-4.1 mini, GPT-4.1 nano (1M token context)
- **GPT-4 Turbo**
- GPT-3.5 Turbo (retiring Feb 2026)
- **GPT-5 mini**, GPT-5 nano, GPT-5 chat
- **o3-mini** (reasoning model)
- GPT-4o audio models (transcribe, TTS)

### Registration required (limited access, ~4 days to 1 week approval):
- **GPT-5** (full) — if you already have o3 access, no new request needed
- **GPT-5.1**, GPT-5.1 codex, GPT-5.1 codex-max
- **GPT-5.2**, GPT-5.2 codex
- o3 streaming

Apply at [aka.ms/oai/access](https://learn.microsoft.com/legal/cognitive-services/openai/limited-access).

---

## Azure OpenAI — How to Get Access

Most Azure OpenAI models are now available to all Azure customers **without a separate registration form**.

1. Create an Azure account (steps above)
2. Go to **Azure AI Foundry** in the Azure Portal
3. **Create an Azure OpenAI resource** in a supported region
4. **Deploy a model** — most models (GPT-4o, GPT-4.1, etc.) deploy immediately
5. For **limited-access models** (GPT-5 full, GPT-5.1, GPT-5.2): submit the [limited access form](https://learn.microsoft.com/legal/cognitive-services/openai/limited-access) and wait ~4 days to 1 week

Additional approval is only needed for modifying content filters or abuse monitoring (requires Microsoft account team relationship).

**Note:** Azure OpenAI and OpenAI API (platform.openai.com) are separate platforms with separate billing and access policies.

---

## AI-Related Free Services

### Always Free (ongoing, within monthly limits)

| Service | Free Limit |
|---|---|
| **Azure AI Search** | 50 MB storage, 10,000 docs, 3 indexes |
| **Azure Language** (sentiment, key phrases, NER) | 5,000 text records/month |
| **AI Bot Service** | 10,000 premium channel msgs + unlimited standard |
| **AI Immersive Reader** | 3M characters/month |
| **Health Bot** | 3,000 messages/month |
| **Speaker Recognition** | 10,000 transactions/month |
| **Speech to Text** | 5 hours/month |
| **Text to Speech** | 500,000 neural characters/month |
| **Speech Translator** | 2M characters/month |
| **Computer Vision** | 5,000 transactions/month |
| **Face API** | 30,000 transactions/month |
| **Azure Machine Learning** | Free (R and Python model development) |

### 12 Months Free

| Service | Free Limit |
|---|---|
| **Azure Document Intelligence** | 500 pages (S0 tier) |
| **AI Custom Vision** | 10,000 predictions + 1 training hour + 2 projects |

### $200 Credit (first 30 days)

The $200 credit can be used for **Azure OpenAI API calls** (GPT-4o, GPT-4.1, etc.), which are not part of the always-free tier but can be fully explored during the trial.

---

## Additional Tips

- **Azure for Students** — separate program: $100 credit, 12 months, no credit card required, needs .edu email
- **Microsoft for Startups Founders Hub** — up to $5,000 in Azure credits for eligible startups
- **Region matters** — not all models are available in all Azure regions; check regional availability before creating resources
- **$200 credit clock starts at signup**, not when you first use a service

---

**Sources:**
- [Azure Free Account](https://azure.microsoft.com/en-us/pricing/purchase-options/azure-account/)
- [Azure Free Services Full List](https://azure.microsoft.com/en-us/pricing/free-services)
- [Azure OpenAI Limited Access Policy](https://learn.microsoft.com/legal/cognitive-services/openai/limited-access)
- [Azure OpenAI Models](https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models)
- [Azure OpenAI What's New](https://learn.microsoft.com/en-us/azure/ai-foundry/openai/whats-new)
- [Azure Sign-Up Troubleshooting](https://learn.microsoft.com/en-us/azure/cost-management-billing/troubleshoot-subscription/troubleshoot-azure-sign-up)
- [Azure for Students](https://azure.microsoft.com/en-us/free/students)

## Submission 25: DigitalOcean Hatch Startup Program - $100,000

- Category: Startup Programs
- Source Link: https://www.digitalocean.com/startups
- Amount: $100,000 in compute credits (12 months) + 3 months free GPU usage
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

The DigitalOcean Hatch program (now branded as "DigitalOcean Startups") gives qualified early-stage startups **up to $100,000 in compute credits over 12 months**, plus **3 months of free GPU usage**, **15 months of Standard-tier support**, marketplace exposure to 600K+ customers, and a welcome kit packed with partner perks from Stripe, Cloudflare, Segment, Notion, and more. AI-native startups are prioritized. The credit amount you receive depends on your partner affiliation — startups referred by recognized accelerators, incubators, or VCs typically get the highest tier. Direct applicants can still apply but may receive less.

---

## Registration (Step-by-Step)

1. Go to [digitalocean.com](https://www.digitalocean.com) and **create a new team account** using your **company domain email** (e.g. you@yourcompany.com). Free email providers like Gmail, Protonmail, or Hotmail are not accepted
2. Add a **valid credit card** to the account during signup
3. Go to [do.co/AIStartups](https://do.co/AIStartups) — this opens the Hatch application form (hosted on Typeform)
4. Fill out the application: company name, website, description of what you are building, funding stage, team size, and select your **affiliated partner organization** (accelerator, incubator, or VC). If you are bootstrapped or unaffiliated, select "Other"
5. Submit the application and wait for a response from the Hatch team
6. If accepted, you receive a welcome email with your credit allocation and access to the **Welcome Kit** with partner perks

**Important:**
- You must be a **new DigitalOcean customer** — no prior accounts or promotional credits
- Your company must have raised a **Series A or less** ($10M cap)
- Service-based businesses (consulting, agencies, web/app dev shops) are **not eligible**
- Admission is at DigitalOcean's discretion and is **not guaranteed**
- If you have questions, email **hatch@digitalocean.com**
- DigitalOcean may request documentation proving eligibility at any time

---

## Credit Structure & Monthly Caps

Hatch credits are distributed as a **fixed monthly allocation** over 12 months — they do not roll over from month to month.

| Detail | Value |
|---|---|
| **Maximum total credits** | $100,000 |
| **Duration** | 12 months (use-it-or-lose-it) |
| **Monthly distribution** | Fixed monthly cap (varies by tier) |
| **Overage** | Charged to your credit card |
| **Monthly hard cap (post-Oct 2020)** | $8,400 above your credit amount triggers a Terms breach |
| **GPU coverage** | Credits do **NOT** cover GPU Droplets |
| **Crypto mining** | Prohibited |
| **Transferable** | No — credits cannot be transferred, assigned, or used for resale |

If you exceed your monthly credit amount, you are billed for the difference. If you exceed $8,400 above your credit amount in a single month, DigitalOcean may terminate your participation.

---

## GPU Access & Pricing

Hatch participants get two GPU-specific benefits:

| Benefit | Details |
|---|---|
| **Free GPU usage** | Up to 3 months at zero cost for qualified startups |
| **Discounted H100 pricing** | ~$1.90/GPU/hour for 12 months (vs. $2.99/hour on-demand standard pricing) |

**Key caveat:** Your regular Hatch compute credits do **not** cover GPU Droplet usage. After the 3-month free period, GPU costs are billed separately to your credit card at the discounted rate. Plan your GPU budget accordingly.

DigitalOcean GPU Droplets use **NVIDIA H100 Tensor Core** GPUs, suitable for model training, fine-tuning, and inference workloads.

---

## Services Covered by Hatch Credits

Hatch credits can be applied to most DigitalOcean infrastructure services:

| Service | Description |
|---|---|
| **Droplets** | CPU-based virtual machines (from $4/mo) |
| **Kubernetes (DOKS)** | Managed Kubernetes clusters (control plane is free) |
| **App Platform** | PaaS for deploying from Git repos (from $5/mo) |
| **Managed Databases** | PostgreSQL, MySQL, MongoDB, Kafka, Redis/Caching |
| **Spaces** | S3-compatible object storage |
| **Volumes** | Block storage |
| **Load Balancers** | Traffic distribution |
| **DigitalOcean Gradient** | AI/ML platform for training, inference, and agent building |
| **Networking** | VPC, firewalls, DNS |

**Not covered by credits:**
- GPU Droplets / H100 products (separate billing)
- Cryptocurrency mining workloads
- Programs consuming excessive network capacity, CPU cycles, or disk I/O

---

## Support & Mentorship

| Benefit | Details |
|---|---|
| **Standard-tier support** | 15 months free (extends 3 months beyond credit period) |
| **Expert office hours** | Monthly webinars and live Q&A sessions |
| **One-on-one meetings** | Access to Product Managers, Solutions Engineers, and Developer Advocates |
| **Marketplace exposure** | Listing opportunity in DigitalOcean Marketplace (600K+ customers) |
| **Co-marketing** | Partner opportunities at DigitalOcean events worldwide |

---

## Welcome Kit — Partner Perks

Accepted startups receive the Hatch Welcome Kit with credits and discounts from partner companies:

| Partner | Perk |
|---|---|
| **Stripe** | 20% off Stripe Atlas (incorporation service) |
| **Cloudflare** | 1 year of Cloudflare Enterprise for free |
| **Segment** | $50,000 in Segment credits for 1 year |
| **HubSpot** | Up to 90% off in your first year |
| **Retool** | 12 months free |
| **Airtable** | $2,000 in credits |
| **Cloud 66** | $500 in credits |
| **QuickNode** | Up to $5,000 in credits |
| **Bolster** | $1,000 in credits |
| **Notion** | Discounted access |
| **GitHub** | GitHub Enterprise seats (details in Welcome Kit) |

Perks are claimed through the Welcome Kit portal after acceptance. Some perks have their own eligibility requirements and expiration windows.

---

## Partner Affiliation & Credit Tiers

The amount of credits you receive depends heavily on how you enter the program:

- **Accelerator/incubator/VC referral** (Y Combinator, Techstars, 100X.VC, Startup India, iCreate, etc.) — highest credit tiers, up to $100,000
- **Direct application ("Other")** — still accepted, but typically lower credit allocation
- **Bootstrapped startups** — eligible, but higher tiers require ecosystem affiliation

If your accelerator or investor is not listed in the application form, select "Other" and describe the relationship. You can also ask your accelerator if they have a DigitalOcean partnership — many do but startups miss it.

---

## Notable Alumni

The Hatch program has been running since 2016 and has supported thousands of startups, including:
- **Hugging Face** — AI model hub
- **Aiven** — managed open-source data infrastructure
- **OneText** — SMS commerce platform
- **Vuukle** — audience engagement platform
- **Ex-human** — empathetic AI agents

---

## Additional Tips

- **Start using credits early** — the 12-month clock starts at acceptance, not at first usage. Unused monthly credits are forfeited each month
- **No extensions** — DigitalOcean does not grant credit extensions under any circumstances
- **One enrollment per company** — you cannot re-enroll or enroll the same business twice
- **GPU free period is separate** — the 3 months of free GPU usage runs independently from your compute credits. After 3 months, GPU usage is billed at the discounted rate (~$1.90/hr)
- **Set billing alerts** — configure alerts in your DigitalOcean dashboard to avoid surprise overages
- **Marketplace listing** — take advantage of the marketplace exposure; it is a unique benefit that most cloud startup programs do not offer
- **Claim all Welcome Kit perks** — many startups forget to redeem partner perks like the $50K Segment credit or free Cloudflare Enterprise year. Check the Welcome Kit immediately after acceptance

---

**Sources:**
- [DigitalOcean Startups (Hatch) Program](https://www.digitalocean.com/startups)
- [DigitalOcean Hatch Application (do.co/AIStartups)](https://do.co/AIStartups)
- [DigitalOcean Announces New Benefits for Hatch Startup Program (Jan 2025)](https://www.businesswire.com/news/home/20250122496359/en/DigitalOcean-Announces-New-Benefits-for-Hatch-Startup-Program)
- [DigitalOcean Hatch Terms of Service](https://www.digitalocean.com/legal/hatch-terms-of-service)
- [DigitalOcean Startups Welcome Kit](https://www.digitalocean.com/startups/welcome-kit)
- [DigitalOcean GPU Droplets Pricing](https://www.digitalocean.com/pricing/gpu-droplets)
- [DigitalOcean Products Overview](https://www.digitalocean.com/products)
- [Hatch Program on CloudCredits.io](https://cloudcredits.io/providers/digitalocean/programs/hatch)
- [Cloudflare Hatch Terms](https://www.cloudflare.com/hatch-terms/)

## Submission 24: DeepSeek - 5 Million Free Tokens for New Users

- Category: AI API Free Tiers
- Source Link: https://platform.deepseek.com/
- Amount: 5 million tokens (~$8.40 value)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

New users who register on the DeepSeek API platform receive **5 million free tokens** (approximately **$8.40 value**) valid for **30 days**. No credit card required. The API gives access to DeepSeek-V3.2 in both chat and reasoning modes via an OpenAI-compatible interface. On top of the API credits, the **chat interface at chat.deepseek.com is completely free and unlimited** for everyone — no account needed. All DeepSeek models are also **open-source under the MIT license**, so you can self-host them at zero cost.

---

## Registration (Step-by-Step)

1. Go to [platform.deepseek.com](https://platform.deepseek.com)
2. Click **"Sign Up"** — you can register with an **email address** or via **GitHub** for quicker access
3. Fill out the registration form (name, email, password) and submit
4. **Verify your email** — click the confirmation link sent to your inbox (usually arrives within a minute)
5. Log into the dashboard and navigate to the **"API Keys"** section
6. Click **"Generate API Key"** — copy and save it immediately, as you will not be able to view it again
7. Your account is automatically credited with **5 million free tokens** — visible in the dashboard balance

**Important:**
- No credit card or payment method is required at any step
- One free token grant per account
- During periods of extremely high demand (viral surges), DeepSeek may temporarily pause new registrations or top-ups — these pauses typically last a few hours to a day and do not affect existing balances
- The 30-day clock starts at registration, not at first API call

---

## Available Models

The DeepSeek API currently serves two model endpoints, both powered by **DeepSeek-V3.2**:

| Model ID | Mode | Context Length | Max Output | Best For |
|---|---|---|---|---|
| **deepseek-chat** | Non-thinking (fast) | 128K tokens | 8K tokens | General chat, coding, content generation |
| **deepseek-reasoner** | Thinking (chain-of-thought) | 128K tokens | 64K tokens | Math, logic, complex reasoning, code analysis |

Both models support JSON output, tool/function calling, and chat prefix completion. The `deepseek-chat` model also supports FIM (Fill-in-the-Middle) completion for code insertion tasks.

---

## API Pricing (After Free Credits)

Once your 5 million free tokens are used up or expire, pricing is pay-as-you-go with no minimums or subscriptions:

| Metric | Price per 1M tokens |
|---|---|
| **Input (cache hit)** | $0.028 |
| **Input (cache miss)** | $0.28 |
| **Output** | $0.42 |

Pricing is **identical for both deepseek-chat and deepseek-reasoner**. For context, this is roughly **95% cheaper than GPT-5** and significantly less than Claude Sonnet 4.

DeepSeek uses aggressive prompt caching — if your requests share common prefixes (system prompts, few-shot examples), subsequent calls automatically get the cache-hit rate, which is **90% cheaper** than a cache miss.

**Granted balance (free tokens) is consumed first** before any topped-up balance.

---

## OpenAI-Compatible API

The DeepSeek API is designed to be a drop-in replacement for OpenAI's API. If you already use the OpenAI SDK, switching is minimal:

- **Base URL:** `https://api.deepseek.com` (or `https://api.deepseek.com/v1` for full compatibility)
- **Authentication:** `Authorization: Bearer YOUR_API_KEY`
- **Endpoint:** `POST /chat/completions`
- Works with the official **OpenAI Python and Node.js SDKs** — just change the base URL and model name

---

## Free Chat Interface (No API Needed)

If you do not need API access, **chat.deepseek.com** provides a completely free web interface:

- **No account required** — open the site and start chatting immediately
- **No message limits** or daily caps — unlimited usage
- **No subscription tiers** — there is no paid "Plus" or "Pro" plan
- Access to **DeepSeek-V3.2** in both Fast mode and **Deep Think** (reasoning) mode
- Supports file uploads, long-context conversations, and web search
- Available on **iOS and Android** apps as well
- Works well in **50+ languages** with strong English and Chinese support

This is ideal for casual use, prototyping prompts, or anyone who does not need programmatic access.

---

## Open-Source Self-Hosting

All DeepSeek models are released under the **MIT license**, allowing unrestricted commercial use:

| Model | Parameters | License | Available On |
|---|---|---|---|
| **DeepSeek-V3.2** | 671B total (37B active per token) | MIT | [Hugging Face](https://huggingface.co/deepseek-ai/DeepSeek-V3.2) |
| **DeepSeek-R1** | 671B total | MIT | [Hugging Face](https://huggingface.co/deepseek-ai/DeepSeek-R1) |
| **R1 Distill Llama 70B** | 70B | MIT | Hugging Face, Ollama |
| **R1 Distill variants** | 1.5B to 70B | MIT | Hugging Face, Ollama |

- Self-hosting means **zero API charges** — you pay only for your own compute
- The distilled models (1.5B–70B) can run on consumer hardware — the smallest fits on a laptop
- Deployment frameworks include SGLang, LightLLM, vLLM, and Ollama
- Over 700 community-built derivative models exist on Hugging Face
- You are free to fine-tune, quantize, distill, and build commercial products

---

## Rate Limits

DeepSeek's API does **not impose fixed rate limits** (no strict RPM or TPM caps). Instead:

- The system uses **dynamic throttling** based on current server load
- During high traffic, requests may be queued rather than rejected
- If a request has not completed within **30 minutes**, the connection is closed
- In practice, most requests complete in seconds under normal conditions

This is unusually generous compared to other providers that enforce strict per-minute quotas.

---

## Additional Tips

- **5M tokens go further than you think** — at the cache-hit rate, 5M input tokens cost only $0.14, so your effective free usage depends heavily on your output volume
- **Use cache-friendly prompt design** — keep system prompts and few-shot examples at the start of your messages to maximize cache hits
- **No auto-billing** — when free credits expire, API calls simply stop working until you top up. There is no surprise charge
- **DeepSeek is a Chinese company** (based in Hangzhou) — data may be processed on servers in China. Consider this for sensitive workloads
- **Third-party hosting** — DeepSeek models are also available through Together.ai, Azure, AWS Bedrock, and other providers if you prefer a different data jurisdiction, though pricing and rate limits will differ
- **The API pricing page changes periodically** — DeepSeek has updated pricing multiple times as they release new model versions. Always check the [official pricing page](https://api-docs.deepseek.com/quick_start/pricing) for current rates

---

**Sources:**
- [DeepSeek API Platform](https://platform.deepseek.com)
- [DeepSeek API Documentation](https://api-docs.deepseek.com/)
- [DeepSeek Models & Pricing](https://api-docs.deepseek.com/quick_start/pricing)
- [DeepSeek Pricing Details (USD)](https://api-docs.deepseek.com/quick_start/pricing-details-usd)
- [DeepSeek Rate Limits](https://api-docs.deepseek.com/quick_start/rate_limit)
- [DeepSeek-V3 on GitHub](https://github.com/deepseek-ai/DeepSeek-V3)
- [DeepSeek-R1 on GitHub](https://github.com/deepseek-ai/DeepSeek-R1)
- [DeepSeek-R1 on Hugging Face](https://huggingface.co/deepseek-ai/DeepSeek-R1)
- [DeepSeek-V3.2 on Hugging Face](https://huggingface.co/deepseek-ai/DeepSeek-V3.2)
- [DeepSeek Chat](https://chat.deepseek.com)

## Submission 23: Databricks for Startups - $50,000 Credits

- Category: Startup Programs
- Source Link: https://www.databricks.com/product/startups
- Amount: $50,000
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Databricks for Startups provides **up to $50,000 in Databricks credits** for VC-funded, data-driven startups, plus **free business-tier technical support** and access to marketing and customer networks. Credits cover Databricks compute (clusters, jobs), SQL warehouses, Mosaic AI Model Serving, notebooks, MLflow, and data engineering workloads on the Databricks Lakehouse Platform. The program is designed for early-stage startups that want to build data-intensive applications without paying full price for enterprise-grade data and AI infrastructure.

---

## Eligibility

- **VC-funded startup** — you must have raised venture capital (angel rounds alone may not qualify)
- **Data-driven** — your product or business should rely on data engineering, analytics, or AI/ML workloads
- **Preferably under $8M in funding** and less than 5 years old (based on third-party reports; Databricks does not publicly confirm exact thresholds)
- **New or free-tier Databricks users** — the deal is available if you have not signed up for Databricks yet or are on the free version. Existing paying customers are unlikely to qualify
- **No specific region restriction** — the program is available globally, though Databricks runs on AWS, Azure, and GCP, so you need access to one of these cloud providers

**Who this is NOT for:**
- Solo developers or hobbyists (use the Databricks Free Edition instead)
- Bootstrapped companies without VC funding
- Established companies already on a paid Databricks plan

---

## How to Apply (Step-by-Step)

1. Go to [databricks.com/product/startups](https://www.databricks.com/product/startups)
2. Click the **application form** button on the page
3. Fill in your startup details — expect to provide company name, website, funding stage, brief description of what you are building, and how you plan to use Databricks
4. You may need to provide **basic financials and a pitch deck** as part of the application
5. Submit and wait for review — applications are **manually reviewed** by the Databricks team
6. If approved, credits are applied to your Databricks account

**Important:**
- There is no public SLA on approval timelines — expect days to a few weeks
- The program does not auto-approve; a human reviews your application
- You do not need a credit card for the application itself, but you will need a cloud provider account (AWS, Azure, or GCP) to actually use Databricks

---

## What the Credits Cover

Databricks credits are measured in **DBUs (Databricks Units)**, a normalized unit of compute power. Your $50K in credits can be applied to:

| Service | What It Does |
|---|---|
| **All-Purpose Compute** | Interactive notebooks, exploratory data analysis, development |
| **Jobs Compute** | Scheduled ETL pipelines, production workloads (cheaper per DBU than all-purpose) |
| **SQL Warehouses** | BI and analytics queries via SQL, Photon-accelerated |
| **Mosaic AI Model Serving** | Deploy ML models as REST APIs with auto-scaling |
| **Lakeflow Pipelines** | Declarative data pipelines (formerly Delta Live Tables) |
| **MLflow** | Experiment tracking, model registry, model deployment |
| **Unity Catalog** | Data governance, access control, lineage tracking |
| **Databricks Assistant** | AI coding assistant built into notebooks |

**Note:** Databricks credits cover the Databricks platform charges only. You still pay your cloud provider (AWS, Azure, or GCP) separately for underlying infrastructure costs like VM instances, storage (S3, ADLS, GCS), and networking. This is a common surprise — budget accordingly.

---

## Databricks Platform Overview

Databricks is a **unified data and AI platform** built on the Lakehouse architecture, combining the best of data lakes and data warehouses. Key capabilities relevant to startups:

- **Data Engineering** — Apache Spark-based ETL, streaming with Structured Streaming, Delta Lake for reliable storage
- **Data Analytics & BI** — SQL warehouses with Photon engine, dashboards, Genie AI-powered analytics
- **Machine Learning** — AutoML, feature engineering, experiment tracking with MLflow, model serving
- **Generative AI** — Mosaic AI for building and deploying LLM-powered applications, foundation model serving (Meta Llama, Mistral, etc.), AI agents
- **Data Governance** — Unity Catalog for centralized access control, lineage, and data quality monitoring

---

## Cloud Provider Options

Databricks runs on all three major cloud providers:

| Cloud | Notes |
|---|---|
| **AWS** | Most mature integration; express signup available |
| **Azure** | Deep Microsoft integration; Azure Databricks is a first-party Azure service |
| **GCP** | Requires Google Cloud Organization and billing account; adds ~$200/month GKE cost per workspace |

Your Databricks startup credits apply regardless of which cloud you choose. Cloud infrastructure costs (VMs, storage) are billed separately by the cloud provider.

---

## Beyond the Credits: Additional Benefits

- **Free business-tier technical support** — normally a paid add-on, included for program participants
- **Go-to-market (GTM) resources** — access to Databricks' marketing channels and customer network to help you reach potential customers
- **Expert technical guidance** — advice from Databricks engineers on architecture and best practices
- **Challenge participation** — eligibility for the **Built-On Databricks Startup Challenge** (over $1M in prizes) and the **Generative AI Startup Challenge**

---

## Related Programs

### Databricks AI Accelerator Program (Higher Tier)

For high-potential early-stage startups, Databricks offers a more comprehensive, **invitation-only** program:

| Feature | Startups Program | AI Accelerator |
|---|---|---|
| **Credits** | Up to $50,000 | Up to $250,000 (funding + credits) |
| **Investment** | None | Angel-style investment from Databricks Ventures |
| **Mentorship** | Technical guidance | Executive sponsorship, 20+ VP-level advisors |
| **VC Network** | No | a]6z, Battery, General Catalyst, Insight, NEA, Menlo, Madrona |
| **Eligibility** | VC-funded startups | Pre-seed/seed, building B2B apps on Databricks |
| **Application** | Open portal | Invitation-only |
| **Post-Program** | — | Entry into Built on Databricks Partner Program |

If you are pre-seed or seed stage and plan to build your core product on Databricks, the AI Accelerator is worth exploring at [databricks.com/databricks-ai-accelerator-program](https://www.databricks.com/databricks-ai-accelerator-program).

### Databricks via JoinSecret ($21,000 Credits)

An alternative route: [JoinSecret](https://www.joinsecret.com/databricks) offers **$21,000 in Databricks credits for 1 year** through their startup deals marketplace. This is less than the direct program's $50K maximum but may have a simpler application process. JoinSecret requires a membership fee.

### Databricks Free Edition (No Credits Needed)

If you do not qualify for the startup program, Databricks offers a **permanently free edition** — no credit card, no expiration. It includes notebooks, SQL editor, dashboards, Lakeflow pipelines, and the Databricks Assistant. Compute is limited (daily quotas, 2X-Small SQL warehouses only), and it is intended for learning and experimentation, not production use.

### Databricks Free Trial ($400 Credits, 14 Days)

Anyone can sign up for a **14-day free trial** with up to $400 in credits. This gives full platform access and is useful for evaluation before applying to the startup program.

---

## Additional Tips

- **Apply before you need it** — the approval process is manual and can take time. Do not wait until you are mid-project
- **Cloud costs are separate** — the $50K covers Databricks platform fees (DBUs) only. Budget for AWS/Azure/GCP compute and storage on top
- **Jobs compute is cheaper** — once past the experimentation phase, move workloads from all-purpose (interactive) clusters to jobs clusters to stretch your credits further
- **Start with the 14-day trial** — use the free trial to validate your architecture and have concrete usage data ready for your startup program application
- **Credit expiration** — Databricks does not publicly disclose the credit validity period for the startup program. Clarify this during the application process
- **No auto-billing trap** — unlike some cloud free tiers, the startup credits do not auto-convert to paid billing without your consent. However, once credits are exhausted, you will need to set up billing to continue

---

**Sources:**
- [Databricks for Startups](https://www.databricks.com/product/startups)
- [Databricks AI Accelerator Program](https://www.databricks.com/databricks-ai-accelerator-program)
- [Databricks AI Accelerator Announcement](https://www.prnewswire.com/news-releases/databricks-launches-ai-accelerator-program-to-scale-the-next-generation-of-ai-apps-and-agents-startups-302560101.html)
- [Databricks for Startups - CloudCredits.io](https://cloudcredits.io/providers/databricks/programs/databricks-for-startups-program)
- [Databricks Free Edition](https://www.databricks.com/learn/free-edition)
- [Databricks Free Trial vs Free Edition](https://docs.databricks.com/aws/en/getting-started/free-trial-vs-free-edition)
- [Databricks Model Serving Pricing](https://www.databricks.com/product/pricing/model-serving)
- [Databricks on JoinSecret](https://www.joinsecret.com/databricks)
- [Databricks Pricing Guide](https://www.chaosgenius.io/blog/databricks-pricing-guide/)

## Submission 22: Cohere Trial API Key - 1,000 Calls/Month

- Category: AI API Free Tiers
- Source Link: https://dashboard.cohere.com/
- Amount: 1,000 API calls per month across all endpoints
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Cohere gives every new account a **free Trial API key** with **1,000 API calls per month** across all endpoints — Chat (Command R+, Command A), Embed 4, Rerank 3.5, and more. No credit card required. The trial key has the same model access as a paid Production key; the only differences are rate limits and a non-commercial use restriction. The monthly cap resets on each calendar month, and the trial never expires as long as your account exists.

---

## Registration (Step-by-Step)

1. Go to [dashboard.cohere.com/welcome/register](https://dashboard.cohere.com/welcome/register)
2. Enter your **email address** and create a **password**, or sign up with **Google** or **GitHub** SSO
3. **Verify your email** by clicking the confirmation link sent to your inbox
4. You are redirected to the Cohere dashboard — a **Trial API key** has already been created for you automatically
5. Navigate to **API Keys** in the sidebar to copy your key

**Important:**
- **No credit card required** — the trial is completely free
- One trial key per account; you can regenerate it from the dashboard
- The trial key works immediately — no approval process or waiting period
- You can sign up with a personal email; no company or organization domain required

---

## Available AI Models

The Trial key gives access to **all Cohere models and endpoints**, identical to a Production key. There are no model-level restrictions.

### Command Family (Text Generation — Chat endpoint)

| Model | Context Length | Notes |
|---|---|---|
| **Command A** (command-a-03-2025) | 256K tokens | Most performant model, 111B parameters |
| **Command A Vision** | 256K tokens | Multimodal — text + image understanding |
| **Command A Reasoning** | 256K tokens | Hybrid reasoning for complex agentic tasks |
| **Command A Translate** | 256K tokens | State-of-the-art machine translation, 23 languages |
| **Command R+** (08-2024) | 128K tokens | Advanced enterprise model |
| **Command R** | 128K tokens | RAG, code generation, workflows |
| **Command R7B** | 128K tokens | Smallest and most cost-effective |

### Embed Family (Embeddings endpoint)

| Model | Dimensions | Notes |
|---|---|---|
| **Embed 4** (Multimodal) | Matryoshka | 128K context, text + image embeddings |
| **Embed English v3.0** | 1,024 | English-only, 512 tokens |
| **Embed English Light v3.0** | 384 | Faster, smaller English model |
| **Embed Multilingual v3.0** | 1,024 | Multiple languages, 512 tokens |

### Rerank Family (Rerank endpoint)

| Model | Notes |
|---|---|
| **Rerank v4.0** | Newest model (December 2025) |
| **Rerank v3.5** | Multilingual, 100+ languages, enhanced reasoning |
| **Rerank English v3.0** | English-only |
| **Rerank Multilingual v3.0** | Multilingual predecessor to 3.5 |

---

## Rate Limits

Trial keys have two layers of throttling: a **monthly cap** of 1,000 calls total and **per-minute limits** that vary by endpoint.

### Per-Minute Rate Limits (Trial Key)

| Endpoint | Calls/Minute |
|---|---|
| **Chat** (`/v2/chat`) | 20 |
| **Embed — Text** (`/v2/embed`) | 100 |
| **Embed — Images** (`/v2/embed`) | 5 |
| **Rerank** (`/v2/rerank`) | 10 |
| **Classify** | 100 |
| **Tokenize** (`/v2/tokenize`) | 100 |
| **Legacy endpoints** | 5 |

### Monthly Cap

- **1,000 API calls/month** across all endpoints combined
- Once you hit the cap, the key returns errors until the next calendar month
- Each API request counts as one call regardless of token count

### Production Key Comparison

| | Trial Key | Production Key |
|---|---|---|
| Monthly limit | 1,000 calls | Unlimited |
| Per-minute limit | 5–100 (varies) | 1,000 all endpoints |
| Cost | Free | Pay-as-you-go |
| Commercial use | Not permitted | Permitted |

---

## Upgrading to Production

If you need higher throughput or want to use Cohere in a production application, you can upgrade to a Production key directly from the dashboard.

1. Go to **API Keys** in the Cohere dashboard
2. You must be an **organization admin** (or ask your admin)
3. **Acknowledge** Cohere's SaaS agreement, terms of service, and model limitations
4. **Add payment information** — billing is pay-as-you-go
5. A Production key is generated immediately (takes under 3 minutes)

**Sensitive use case review:** If your application involves a sensitive use case (healthcare, legal, employment decisions, etc.), your Production key may remain rate-limited like a trial key until Cohere's safety team manually reviews and approves your use case.

### Production Pricing (for reference)

| Model | Input | Output |
|---|---|---|
| **Command A** | $2.50 / 1M tokens | $10.00 / 1M tokens |
| **Command R+** | $2.50 / 1M tokens | $10.00 / 1M tokens |
| **Command R** | $0.15 / 1M tokens | $0.60 / 1M tokens |
| **Command R7B** | $0.0375 / 1M tokens | $0.15 / 1M tokens |
| **Embed 4 (text)** | $0.12 / 1M tokens | — |
| **Embed 4 (image)** | $0.47 / 1M tokens | — |
| **Rerank 3.5** | $2.00 / 1K searches | — |

Billing is issued at the end of each calendar month or when outstanding balance reaches $250, whichever comes first.

---

## Non-Commercial Use Restriction

The Trial key is explicitly **not permitted for production or commercial use**. This means:
- You can use it for **prototyping**, **learning**, **research**, and **proof-of-concept** projects
- You **cannot** serve it to end users in a commercial product
- You **cannot** use it in revenue-generating applications
- Violating this policy may result in account suspension

---

## Additional Tips

- **Command A Reasoning has special limits** — even Production keys are rate-limited like trial keys (1,000 calls/month) for newer model variants like Command A Reasoning. Contact sales@cohere.com for higher limits on these models
- **SDKs available** — official SDKs for Python (`cohere`), TypeScript/Node.js, Go, and Java. Install with `pip install cohere` and pass your trial key
- **Cohere on cloud marketplaces** — Cohere models are also available on Amazon Bedrock, Amazon SageMaker, and Oracle Cloud Infrastructure, each with their own pricing and free tiers
- **Playground** — the Cohere dashboard includes a built-in playground for testing prompts without writing code
- **Aya models** — multilingual models (Aya Expanse, Aya Vision) are also accessible via the trial key for multilingual and multimodal experimentation
- **No auto-billing risk** — since no credit card is required for the trial, there is zero risk of accidental charges. You must explicitly upgrade and add payment info to incur any costs

---

**Sources:**
- [Cohere Rate Limits Documentation](https://docs.cohere.com/docs/rate-limits)
- [Cohere Pricing](https://cohere.com/pricing)
- [Cohere Pricing Documentation](https://docs.cohere.com/docs/how-does-cohere-pricing-work)
- [Cohere Models Overview](https://docs.cohere.com/docs/models)
- [Cohere Going Live Guide](https://docs.cohere.com/docs/going-live)
- [Cohere FAQs](https://docs.cohere.com/docs/cohere-faqs)
- [Cohere Usage Policy](https://docs.cohere.com/docs/usage-policy)
- [Cohere Trial Key Summary (CodeNote)](https://codenote.net/en/posts/cohere-trial-api-key-pricing-and-limits/)
- [Cohere API Pricing 2026 (MetaCTO)](https://www.metacto.com/blogs/cohere-pricing-explained-a-deep-dive-into-integration-development-costs)

## Submission 21: Cloudflare Workers AI - 10,000 Free Neurons/Day

- Category: AI API Free Tiers
- Source Link: https://developers.cloudflare.com/workers-ai
- Amount: 10,000 Neurons/day (daily reset at 00:00 UTC)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Cloudflare Workers AI gives you **10,000 free Neurons per day** to run AI models on Cloudflare's global edge network — no credit card required, no time limit. Neurons measure GPU compute across all model types (LLMs, image generation, embeddings, speech-to-text, text-to-speech, and more). With the free daily allowance you can generate roughly **1,300 LLM responses**, **8,300 image classifications**, or **12,500 embeddings**. The limit resets every day at 00:00 UTC. Workers Free plan users hit a hard rate limit after 10,000 neurons; Workers Paid plan users ($5/month) get the same 10,000 free neurons plus overflow at $0.011 per 1,000 neurons.

---

## Registration (Step-by-Step)

1. Go to [dash.cloudflare.com/sign-up](https://dash.cloudflare.com/sign-up/workers-and-pages) and create a free Cloudflare account (email + password, no credit card)
2. Verify your email address
3. In the dashboard, navigate to **Workers & Pages**
4. You now have three ways to start using Workers AI:

### Option A: Dashboard (no CLI, easiest)
5. Click **Create application** and select the **"LLM Chat App"** template
6. Name your Worker and click **"Create and deploy"**
7. Your app is live on a `workers.dev` subdomain — click **"Edit Code"** to modify

### Option B: REST API (no Worker needed)
5. In the dashboard, go to **Workers AI** and click **"Use REST API"**
6. Click **"Create a Workers AI API Token"** — review the prefilled permissions (`Workers AI - Read` and `Workers AI - Edit`) and generate the token
7. Copy your **Account ID** from the same page
8. Make your first call:
```bash
curl https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/run/@cf/meta/llama-3.1-8b-instruct \
  -H 'Authorization: Bearer {API_TOKEN}' \
  -d '{"prompt": "Hello, what can you do?"}'
```

### Option C: Wrangler CLI (for developers)
5. Install Node.js (16.17+) and run `npm create cloudflare@latest`
6. Add an `[ai]` binding to your `wrangler.toml`
7. Call `env.AI.run()` in your Worker code
8. Deploy with `wrangler deploy`

**Important:**
- No credit card is required for the free plan
- Using Workers AI locally via `wrangler dev` still consumes your Cloudflare account's neuron quota
- The API token needs both Read and Edit permissions for Workers AI

---

## OpenAI-Compatible API

Workers AI supports **OpenAI-compatible endpoints**, so you can use the standard OpenAI SDK by swapping the base URL:

| Endpoint | URL |
|---|---|
| **Chat Completions** | `https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/v1/chat/completions` |
| **Text Completions** | `https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/v1/completions` |
| **Embeddings** | `https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/v1/embeddings` |

All standard OpenAI parameters work: `temperature`, `max_tokens`, `top_p`, `frequency_penalty`, `presence_penalty`. Many models also support **function calling**, **streaming**, and **batch processing**.

---

## Available AI Models

Workers AI hosts **100+ open-source models** across multiple task types. All models run on Cloudflare's edge GPUs — no setup, no cold starts for popular models.

### Text Generation (LLMs)

| Model | Author | Notes |
|---|---|---|
| **gpt-oss-120b** | OpenAI | Powerful reasoning, agentic tasks, general purpose |
| **gpt-oss-20b** | OpenAI | Lower latency, specialized use cases |
| **Llama 4 Scout 17B** | Meta | Mixture-of-experts, function calling |
| **Llama 3.3 70B Instruct (FP8)** | Meta | Quantized, batch + function calling |
| **Llama 3.1 8B Instruct** | Meta | Multilingual dialogue, multiple quantizations available |
| **Llama 3.2 11B Vision** | Meta | Vision + text, image reasoning |
| **Qwen3 30B-A3B (FP8)** | Qwen | Reasoning, instruction-following, agent capabilities |
| **QwQ 32B** | Qwen | Reasoning model for complex problems |
| **Qwen 2.5 Coder 32B** | Qwen | Code-specialized |
| **Gemma 3 12B IT** | Google | Multimodal, 128K context, 140+ languages |
| **Mistral Small 3.1 24B** | Mistral | Vision-capable, 128K tokens |
| **DeepSeek R1 Distill Qwen 32B** | DeepSeek | Distilled reasoning model |
| **Granite 4.0-H Micro** | IBM | Agentic tasks, instruction following |
| **Hermes 2 Pro Mistral 7B** | Nous Research | Function calling + JSON mode |

### Image Generation

| Model | Author | Notes |
|---|---|---|
| **FLUX.2 Klein 9B** | Black Forest Labs | Enhanced quality, generation + editing |
| **FLUX.2 Klein 4B** | Black Forest Labs | Ultra-fast distilled |
| **FLUX.2 Dev** | Black Forest Labs | Highly realistic, multi-reference support |
| **FLUX.1 Schnell** | Black Forest Labs | 12B parameter, fast generation |
| **Lucid Origin** | Leonardo | Strong prompt adherence, text rendering |
| **Phoenix 1.0** | Leonardo | Coherent text in images |
| **Stable Diffusion XL Lightning** | ByteDance | Lightning-fast 1024px generation |

### Embeddings

| Model | Author | Notes |
|---|---|---|
| **EmbeddingGemma 300M** | Google | 100+ languages |
| **Qwen3 Embedding 0.6B** | Qwen | Text embedding and ranking |
| **BGE-M3** | BAAI | Multi-lingual, multi-granularity |
| **BGE Large/Base/Small EN v1.5** | BAAI | English embeddings (384–1024 dims) |

### Speech-to-Text

| Model | Author | Notes |
|---|---|---|
| **Nova 3** | Deepgram | High-accuracy transcription |
| **Flux** | Deepgram | Conversational speech recognition for voice agents |
| **Whisper Large v3 Turbo** | OpenAI | Multilingual ASR and translation |
| **Whisper** / **Whisper Tiny EN** | OpenAI | General-purpose speech recognition |

### Text-to-Speech

| Model | Author | Notes |
|---|---|---|
| **Aura 2** (EN/ES) | Deepgram | Context-aware, natural pacing, real-time |
| **Aura 1** | Deepgram | Natural prosody |
| **MeloTTS** | MyShell AI | Multi-lingual TTS |

### Other Tasks

- **Image Classification:** ResNet-50 (Microsoft)
- **Object Detection:** DETR ResNet-50 (Facebook)
- **Translation:** M2M-100 1.2B (Meta), IndicTrans2 (AI4Bharat)
- **Summarization:** BART Large CNN (Facebook)
- **Text Classification / Reranking:** BGE Reranker, DistilBERT
- **Image-to-Text:** LLaVA 1.5 7B, UForm Gen2
- **Safety:** Llama Guard 3 8B (prompt/response classification)
- **Voice Activity Detection:** Pipecat Smart Turn v2

---

## Understanding Neurons & What You Get for Free

Neurons abstract GPU compute across model types. Here is what **10,000 neurons** (one day's free allowance) roughly translates to:

| Task | Approximate Free Daily Volume |
|---|---|
| **LLM responses** (e.g., Llama 3.1 8B) | ~1,300 responses |
| **Image classifications** | ~8,300 classifications |
| **Text embeddings** | ~12,500 embeddings |
| **Image generation** (FLUX.1 Schnell, 512x512) | ~2,000 images |

Neuron cost varies significantly by model size. Smaller models (Llama 3.2 1B) consume far fewer neurons per request than large ones (Llama 3.3 70B or DeepSeek R1 32B). Choose your model strategically to maximize free usage.

### Per-Model Token Pricing (examples)

| Model | Input (per M tokens) | Output (per M tokens) |
|---|---|---|
| Llama 3.2 1B Instruct | $0.027 | $0.201 |
| Llama 3.2 3B Instruct | $0.051 | $0.335 |
| Llama 3.1 8B Instruct (FP8) | $0.045 | $0.384 |
| Llama 3.1 70B Instruct (FP8) | $0.293 | $2.253 |

---

## Rate Limits

| Plan | Daily Neurons | After Limit | Per-Model Request Limit |
|---|---|---|---|
| **Workers Free** | 10,000 | Hard block (error returned) | 1,500–3,000 RPM depending on model |
| **Workers Paid** ($5/mo) | 10,000 free | $0.011 per 1,000 additional neurons | Higher RPM limits |

The daily neuron quota resets at **00:00 UTC**. Per-model rate limits (requests per minute) are separate from the neuron budget and vary by model — check the docs for each model's specific RPM cap.

---

## Complementary Cloudflare Services

Workers AI integrates with other Cloudflare developer platform products on the free plan:

| Service | Free Tier |
|---|---|
| **Workers** (compute) | 100,000 requests/day, 10ms CPU time per invocation |
| **Vectorize** (vector database) | 5M stored vectors, 30M queried vectors/month |
| **AI Gateway** (observability) | Unlimited requests — caching, rate limiting, logging, model fallback |
| **D1** (SQL database) | 5 GB storage, 5M rows read/day |
| **R2** (object storage) | 10 GB storage, 1M Class A ops/month |
| **KV** (key-value store) | 100,000 reads/day |

This means you can build a full-stack AI application (LLM + embeddings + vector search + storage + caching) entirely on Cloudflare's free tier.

---

## Additional Tips

- **No credit card required** — the free plan is genuinely free with no payment info needed. The $5/month paid plan only matters if you need to exceed 10,000 neurons/day
- **LoRA support** — you can bring your own fine-tuned LoRA adapters and run them on supported base models (Llama, Mistral, Gemma) at no extra cost beyond normal neuron usage
- **Edge latency** — models run on Cloudflare's global network, so inference happens close to your users. Cold starts are minimal for popular models
- **AI Gateway for observability** — route your requests through AI Gateway (free) to get logging, caching (avoid re-running identical prompts), and automatic retries/fallback across providers
- **Neuron tracking** — monitor your usage in the Cloudflare dashboard under Workers AI analytics. The dashboard shows usage in both neurons and conventional units (tokens, seconds, images)
- **Cloudflare is transitioning pricing** from neurons to per-unit pricing (tokens, audio seconds, image tiles) for clarity, but neurons remain the underlying billing metric
- **Startup / enterprise programs** — Cloudflare offers a [Workers Launchpad](https://www.cloudflare.com/lp/workers-launchpad/) program for startups building on their platform, which may include additional credits
- **Model availability** — not every model is available in every Cloudflare data center. Popular models have wider distribution; less-used models may route to fewer locations with slightly higher latency

---

**Sources:**
- [Workers AI Overview](https://developers.cloudflare.com/workers-ai/)
- [Workers AI Pricing](https://developers.cloudflare.com/workers-ai/platform/pricing/)
- [Workers AI Models Catalog](https://developers.cloudflare.com/workers-ai/models/)
- [Get Started — REST API](https://developers.cloudflare.com/workers-ai/get-started/rest-api/)
- [Get Started — Dashboard](https://developers.cloudflare.com/workers-ai/get-started/dashboard/)
- [Get Started — Workers & Wrangler](https://developers.cloudflare.com/workers-ai/get-started/workers-wrangler/)
- [OpenAI Compatible API Endpoints](https://developers.cloudflare.com/workers-ai/configuration/open-ai-compatibility/)
- [Workers AI GA Announcement](https://blog.cloudflare.com/workers-ai-ga-huggingface-loras-python-support/)
- [Workers AI Updated Pricing Changelog](https://developers.cloudflare.com/changelog/2025-02-20-updated-pricing-docs/)
- [OpenAI Open Models on Workers AI](https://developers.cloudflare.com/changelog/2025-08-05-openai-open-models/)

## Submission 20: Cloudflare for Startups - Seed ($100,000)

- Category: Startup Programs
- Source Link: https://www.cloudflare.com/forstartups
- Amount: $100,000
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Seed-funded startups affiliated with a Cloudflare-approved VC or accelerator can receive **$100,000 in credits** covering the full Cloudflare Developer Platform and up to **three enterprise-level domains** with CDN, WAF, DDoS protection, DNS, and Zero Trust features. Credits are valid for **1 year or until fully consumed**, whichever comes first. Subject to per-service caps: **$10K on R2/Cache Reserve** and **$50K on Workers AI**. No credit card charge until credits run out.

---

## Registration (Step-by-Step)

1. Go to [cloudflare.com/forstartups](https://www.cloudflare.com/forstartups/)
2. **Create a Cloudflare account** if you don't already have one (free signup, credit card required on file)
3. Add at least one domain to your Cloudflare account and update your name servers
4. Click **"Apply"** on the Startup Program page
5. Fill out the application form — include details about what you're building, your team, and your funding stage
6. **Select your accelerator or VC** from the dropdown menu. If your program isn't listed, select "Referred by" and include details in the comments field
7. Link to a **landing page, pitch deck, or article** that clearly describes your startup and what you're building
8. Submit the form — you'll receive an automated confirmation email
9. Once approved, credits are applied to your Cloudflare account

**Important:**
- Credits can only be allocated to **one account per applicant**
- Once credits are exhausted or 1 year passes, the credit card on file will be charged for any further usage
- Not available to **existing Cloudflare enterprise customers**
- Cloudflare may consider requests for additional credits on a case-by-case basis — explain your needs in the comments section
- For questions, contact **startups@cloudflare.com**

---

## Eligibility Requirements

The $100,000 (Seed) tier builds on the lower tiers. You must meet **all** of these criteria:

| Requirement | Details |
|---|---|
| **Product** | Building a software-based product or service |
| **Company age** | Founded within the last 5 years |
| **Email & website** | Valid matching email address and active website |
| **LinkedIn** | Active LinkedIn profile for the founder |
| **Funding** | Raised between **$1M and $5M** in capital |
| **VC/Accelerator** | Must be affiliated with one of **250+ Cloudflare-approved VC or accelerator partners** |

If your accelerator or VC is not on the approved list, email **startups@cloudflare.com** with your perk manager's details, and Cloudflare will reach out to them to establish a partnership.

---

## Program Tiers (All Four)

Cloudflare offers four credit tiers. The $100K Seed tier is the third:

| Tier | Credits | Target Stage | Key Extra Requirement |
|---|---|---|---|
| Bootstrapped | **$5,000** | Stealth / early-stage | Software product, founded <5 years, valid email |
| Up-and-Coming | **$25,000** | Early adopters, PMF | Active LinkedIn, funded up to $1M |
| **Seed** | **$100,000** | Raised capital, high growth | Funded $1M–$5M, approved VC/accelerator |
| Scaling | **$250,000** | Mission-critical AI | Tier 1 VC/accelerator, AI company, or Workers Launchpad |

Bootstrapped startups with under $50K in funding can use promo code **"BOOTSTRAPPED"** to access the $5K tier.

---

## Covered Services

### Developer Platform (usage-based, covered by credits)

| Service | What It Does |
|---|---|
| **Workers** | Serverless compute at the edge |
| **Pages** | Frontend hosting and deployment |
| **R2** | Object storage with zero egress fees (capped at **$10K**) |
| **D1** | Serverless SQL database |
| **KV** | Key-value storage |
| **Durable Objects** | Stateful serverless objects |
| **Workers AI** | Inference on 50+ open-source AI models (capped at **$50K**) |
| **Queues** | Message queues |
| **Workflows** | Orchestration engine |
| **Stream** | Video delivery and encoding |
| **Images** | Image optimization and transformation |
| **Argo Smart Routing** | Intelligent traffic routing |

### Enterprise-Level Features (up to 3 domains)

| Feature | Description |
|---|---|
| **CDN** | Global content delivery network |
| **DNS** | Enterprise DNS with DNSSEC |
| **WAF** | Advanced Web Application Firewall |
| **DDoS Protection** | Unlimited, unmetered DDoS mitigation |
| **Zero Trust** | Network security as-a-service |
| **SSL/TLS** | Custom SSL certificates |
| **Cache Reserve** | Persistent cache storage (capped at **$10K** combined with R2) |
| **Rate Limiting** | Advanced rate limiting rules |
| **Load Balancing** | Global load balancing |

---

## Credit Caps & Exclusions

### Per-Service Credit Caps

| Service | Credit Cap |
|---|---|
| **R2 + Cache Reserve** | $10,000 combined |
| **Workers AI** | $50,000 |
| **All other services** | No sub-cap (uses remaining credit pool) |

### Not Included

- **Registrar** — domain registration purchases must be paid in full separately
- **Account management** — no dedicated account manager
- **24/7 enterprise phone support** — not included
- **SLA** — no service level agreement

---

## Workers AI Models

Workers AI provides serverless inference on **50+ open-source models** running on Cloudflare's edge network. Notable models available include:

| Category | Example Models |
|---|---|
| **Large Language Models** | Llama 4 Scout 17B, Mistral Small 3.1 24B, Qwen 2.5 Coder 32B, OpenAI GPT-OSS 120B/20B |
| **Image Generation** | FLUX.2 dev, FLUX.2 klein 9B/4B, Leonardo Phoenix |
| **Speech-to-Text** | Whisper, Deepgram Nova 3 |
| **Text-to-Speech** | Deepgram Aura 2 |
| **Embeddings** | Google EmbeddingGemma 300M, BAAI BGE Large |
| **Image Classification** | Various vision models |

Workers AI is priced at **$0.011 per 1,000 Neurons** (Cloudflare's billing unit). With the $50K Workers AI cap, this gives substantial room for inference workloads. The full model catalog is at [developers.cloudflare.com/workers-ai/models](https://developers.cloudflare.com/workers-ai/models/).

---

## Workers Launchpad Program

If your startup qualifies for the $250K tier or you want additional support beyond credits, consider applying to the **Workers Launchpad** — a complementary multi-month program that includes:

- Hands-on **bootcamp sessions**
- **Solutions Architect** office hours
- Introductions to **venture capital firms**
- Opportunity to present at **Demo Day**

Workers Launchpad participation automatically qualifies you for the $250K credit tier.

---

## What Happens When Credits Run Out

1. **Credits exhausted or 1 year passes** (whichever first) — your Startup Plan ends
2. **Usage-based services** (Workers, R2, D1, etc.) transition to standard pay-as-you-go billing against the credit card on file
3. **Enterprise domain features** revert to your account's plan level (free, Pro, or Business)
4. **No auto-cancellation warning** — set up billing alerts in the Cloudflare dashboard to track credit consumption
5. You can request additional credits on a case-by-case basis by contacting startups@cloudflare.com

---

## Additional Tips

- **AI startups get a boost** — if you're building with LLMs or AI, mention this in your application. AI-focused startups may qualify for the $250K tier even without a Tier 1 VC affiliation
- **Cloudflare's free tier is generous on its own** — Workers (100K requests/day), R2 (10 GB/month), D1, KV, and Pages all have permanent free tiers that persist after the startup plan ends
- **Zero egress fees on R2** — unlike AWS S3, Cloudflare R2 charges nothing for data egress, making it attractive for data-heavy startups even post-credits
- **4,000+ startups** have already received credits since the program launched in 2024
- **Physical co-working hubs** — Cloudflare is opening offices in San Francisco, Austin, London, and Lisbon for startups and builders to use
- **Non-profits** — Cloudflare has a separate program for non-profit and civil society organizations
- **One account per applicant** — credits cannot be split across multiple Cloudflare accounts

---

**Sources:**
- [Cloudflare for Startups](https://www.cloudflare.com/forstartups/)
- [Startup Program Update: Empowering Every Stage](https://blog.cloudflare.com/expanding-cloudflares-startup-program/)
- [Startup Program Revamped: Up to $250,000 in Credits](https://blog.cloudflare.com/startup-program-250k-credits/)
- [Startup Plan v2.0 Blog Post](https://blog.cloudflare.com/startup-program-v2/)
- [Workers AI Models](https://developers.cloudflare.com/workers-ai/models/)
- [Workers & Pages Pricing](https://www.cloudflare.com/plans/developer-platform/)
- [R2 Pricing](https://developers.cloudflare.com/r2/pricing/)
- [Workers AI Pricing](https://developers.cloudflare.com/workers-ai/platform/pricing/)
- [Cloudflare New Hubs for Startups](https://blog.cloudflare.com/new-hubs-for-startups/)

## Submission 19: Cloudflare for Startups - High-Growth ($250,000)

- Category: Startup Programs
- Source Link: https://www.cloudflare.com/forstartups
- Amount: $250,000
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Cloudflare's top-tier startup program provides **$250,000 in credits valid for 1 year** to high-growth startups building on the Cloudflare Developer Platform. This is the highest of four tiers ($5K / $25K / $100K / $250K) and targets startups affiliated with Tier 1 VCs or accelerators, participants in the Workers Launchpad program, or companies building mission-critical AI applications. Credits cover virtually all pay-as-you-go Cloudflare products, and the plan includes up to **three Enterprise-level domains** with CDN, DDoS protection, DNS, WAF, and Zero Trust. Starting January 2026, participants also get access to free **Cloudflare Startup Hubs** (coworking spaces) in San Francisco, Austin, London, and Lisbon.

---

## Registration (Step-by-Step)

1. Go to [cloudflare.com/forstartups](https://www.cloudflare.com/forstartups/)
2. **Create a free Cloudflare account** if you don't have one, and add a valid credit card
3. Click the application link and **fill out the startup application form** — include details about your product, team, funding stage, and any VC/accelerator affiliation
4. If you are part of a Tier 1 VC/accelerator network, make sure to mention this explicitly in the application — it qualifies you for the $250K tier
5. If you are a **Workers Launchpad** participant or alumnus, mention this as well
6. Submit the form — you will receive an automated confirmation email
7. Wait for a review email confirming your acceptance and credit activation

**Important:**
- Acceptance is at Cloudflare's **sole discretion** — there is no guaranteed approval
- Credits are allocated to **one account per applicant** only
- Credits are **non-transferable, non-refundable, and non-redeemable** for cash
- You cannot sell, purchase, or barter credits
- The Startup Plan does **not** include Cloudflare Registrar — domain purchases must be paid separately

---

## Eligibility for the $250,000 Tier

The $250K tier has the strictest requirements. You must meet the base criteria **plus** at least one of the premium qualifiers:

### Base Criteria (all tiers)
- Building a **software-based product or service**
- Founded within the **last 5 years**
- Have a valid website, LinkedIn profile, and matching email address
- Raised between **$50,000 and $5,000,000** in funding

### Premium Qualifiers (need at least one)
- Affiliated with a **Tier 1 VC or accelerator** from Cloudflare's 250+ approved partner network
- Participating in the **Workers Launchpad Program**
- Building a **mission-critical AI application** with strategic potential

---

## All Four Program Tiers

| Tier | Credits | Target Stage | Key Requirements |
|---|---|---|---|
| **Tier 1** | $5,000 | Bootstrapped / stealth | Software product, founded <5 years, valid email |
| **Tier 2** | $25,000 | Up-and-coming | Tier 1 criteria + LinkedIn + up to $1M funding |
| **Tier 3** | $100,000 | Seed-funded | $1M–$5M funding OR membership with approved VC/accelerator |
| **Tier 4** | $250,000 | High-growth / AI | Tier 1 VC network, Workers Launchpad, or strategic AI company |

Bootstrapped startups that haven't raised $50K can use promo code **"BOOTSTRAPPED"** to apply for a lower tier.

---

## What the Credits Cover

Credits apply to the vast majority of Cloudflare products on pay-as-you-go plans:

### Developer Platform
| Product | Description |
|---|---|
| **Workers** | Serverless compute at the edge (200+ cities) |
| **Pages** | Full-stack web application hosting |
| **R2** | Egress-free object storage (S3-compatible) |
| **D1** | Serverless SQL database |
| **Durable Objects** | Stateful serverless computing |
| **Workers KV** | Global key-value storage |
| **Queues** | Message queuing |
| **Hyperdrive** | Database connection pooling |
| **Workers AI** | Serverless AI inference (50+ models) |
| **AI Gateway** | Proxy, cache, and monitor AI API calls |
| **Vectorize** | Vector database for embeddings |

### Performance & Security
| Product | Description |
|---|---|
| **Argo Smart Routing** | Optimized traffic routing |
| **Load Balancing** | Traffic distribution |
| **Cache Reserve** | Extended cache storage |
| **Advanced WAF** | Web application firewall |
| **Rate Limiting** | Request throttling |
| **Zero Trust** | Secure access for teams |

### Enterprise Features (up to 3 domains)
- CDN with full optimization
- DDoS protection (unmetered)
- Enterprise DNS
- WAF managed rulesets
- Zero Trust access

---

## Credit Sub-Limits

While the total pool is $250,000, certain services have hard caps:

| Service | Maximum Credit Usage |
|---|---|
| **R2 + Cache Reserve** | $10,000 |
| **Workers AI** | $50,000 |
| **All other products** | Remaining balance (up to $190,000) |

These sub-limits apply regardless of your total credit amount.

---

## Workers AI — Available Models & Pricing

Workers AI provides serverless inference on 50+ open-source models. Pricing uses **Neurons** (a unit of GPU compute):

- **Free allocation:** 10,000 Neurons/day (available to everyone, even outside the startup program)
- **Paid rate:** $0.011 per 1,000 Neurons beyond the free allocation

### Selected Model Pricing

| Model | Input | Output |
|---|---|---|
| **Llama 3.2 1B Instruct** | $0.027/M tokens | $0.201/M tokens |
| **Llama 3.2 3B Instruct** | $0.051/M tokens | $0.335/M tokens |
| **Llama 3.1 8B Instruct (FP8)** | $0.045/M tokens | $0.384/M tokens |
| **Llama 3.2 11B Vision** | $0.049/M tokens | $0.676/M tokens |
| **Llama 3.3 70B Instruct (FP8)** | $0.293/M tokens | $2.253/M tokens |

Models span text generation, image classification, object detection, embeddings, speech-to-text, and image generation. All inference runs on Cloudflare's global network (200+ cities). The API is **OpenAI-compatible**.

---

## Workers Launchpad Program

The Workers Launchpad is a separate but complementary program that gives participants automatic eligibility for the $250K startup tier:

- **Funding pool:** $2 billion+ through partnerships with 40+ VC firms (including Kleiner Perkins, Greylock, NEA, Bessemer, Lightspeed, and others)
- **Eligibility:** Any startup built on Cloudflare Workers, funded up to Series B, founded in the last 5 years
- **Application:** Takes about 5 minutes at [cloudflare.com/lp/workers-launchpad](https://www.cloudflare.com/lp/workers-launchpad/)
- **Current status:** Cohort #7 applications are open (deadline: March 13, 2026)
- **What you get:** VC pitch opportunities, virtual boot camp, technical guidance, Demo Day showcase, and automatic Startup Plan enrollment

**Note:** Cloudflare does not directly provide funding or make investment decisions. VC partners independently evaluate and fund startups.

---

## Cloudflare Startup Hubs

Starting **January 2026**, Cloudflare is opening its office spaces as free coworking hubs for startup program participants:

| City | Office Highlights |
|---|---|
| **San Francisco** | Cloudflare HQ |
| **Austin** | High-ceilinged collaboration areas |
| **London** | European hub |
| **Lisbon** | Quiet library zone, waterfront location |

### How to Access
1. Be enrolled in the Cloudflare for Startups Program
2. Workers Launchpad participants/alumni get **priority access**
3. Wait for an email invitation with an access request form
4. Register your visit at least **3 business days in advance**
5. No mandatory talks or obligations — just free workspace

Cloudflare plans to expand to additional locations in the future.

---

## What Happens When Credits Expire

- Credits expire after **1 year** or when fully consumed, whichever comes first
- After expiration, your account transitions to standard pay-as-you-go billing
- Enterprise domain features revert to your current plan level
- There is no automatic renewal of startup credits
- You can reapply if you still meet eligibility criteria, though approval is not guaranteed

---

## Additional Tips

- **Track your credit usage** in the Cloudflare dashboard — with sub-limits on R2 ($10K) and Workers AI ($50K), you could hit a cap before exhausting the full $250K
- **Apply to Workers Launchpad simultaneously** — it increases your chances for the $250K tier and gives access to VC introductions
- **AI Gateway is free** for all Cloudflare accounts and does not consume credits — use it to proxy, cache, and monitor calls to external AI APIs (OpenAI, Anthropic, etc.)
- **R2 has zero egress fees** — a major cost advantage over AWS S3 or Google Cloud Storage, and particularly valuable for AI model serving
- **Cloudflare's free plan** already includes generous limits (100K Workers requests/day, 10GB R2 storage, 5M D1 row reads/day) — evaluate whether you even need the startup program for your current scale
- **Over $370 million in credits** have been awarded to 4,000+ startups, with participants collectively raising over $5 billion in funding
- The program is **global** — no region restrictions on eligibility or service access

---

**Sources:**
- [Cloudflare for Startups](https://www.cloudflare.com/forstartups/)
- [Startup Program Revamped: Up to $250,000 in Credits](https://blog.cloudflare.com/startup-program-250k-credits/)
- [Startup Program Update: Empowering Every Stage](https://blog.cloudflare.com/expanding-cloudflares-startup-program/)
- [Come Build With Us: Cloudflare's New Hubs for Startups](https://blog.cloudflare.com/new-hubs-for-startups/)
- [Workers Launchpad Funding Program](https://www.cloudflare.com/lp/workers-launchpad/)
- [Workers Launchpad Cohort #6 Lookback](https://blog.cloudflare.com/workers-launchpad-006/)
- [Workers AI Pricing](https://developers.cloudflare.com/workers-ai/platform/pricing/)
- [Workers AI Models](https://developers.cloudflare.com/workers-ai/models/)
- [Cloudflare Developer Platform Pricing](https://www.cloudflare.com/plans/developer-platform/)
- [R2 Pricing](https://developers.cloudflare.com/r2/pricing/)
- [D1 Pricing](https://developers.cloudflare.com/d1/platform/pricing/)

## Submission 18: Cloudflare for Startups - Growth ($25,000)

- Category: Startup Programs
- Source Link: https://www.cloudflare.com/forstartups
- Amount: $25,000
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Funded startups receive **$25,000 in Cloudflare credits valid for 1 year** covering the full Developer Platform (Workers, R2, D1, Pages, Durable Objects, Queues) plus **up to three Enterprise-level domains** with advanced WAF, DDoS protection, and CDN. This is the second tier of four in Cloudflare's startup program ($5K / $25K / $100K / $250K). You must be a software-based startup founded in the last 5 years with an active LinkedIn profile and funding up to $1M. No credit card hold or identity verification beyond the standard free Cloudflare account signup. Acceptance is at Cloudflare's discretion.

---

## Registration (Step-by-Step)

1. Go to [cloudflare.com/forstartups](https://www.cloudflare.com/forstartups/)
2. **Create a free Cloudflare account** if you do not already have one (email + password, add a valid credit card)
3. Click **"Apply"** on the startup program page
4. Fill out the application form with:
   - Company name, website, and matching company email
   - LinkedIn profile URL (required for the $25K tier)
   - Brief description of what you are building
   - Funding stage and amount raised (up to $1M for this tier)
   - Accelerator or VC partner affiliation (bonus, not required)
5. Submit the form -- you will receive an automated confirmation email
6. Wait for review -- acceptance is at Cloudflare's sole discretion, typically a few business days
7. Once approved, credits are applied to your Cloudflare account and Enterprise domains are provisioned

**Important:**
- This is for **new, non-paying Cloudflare users** only -- existing Enterprise customers are not eligible
- Agencies and service companies are **not eligible**
- If you have not raised at least $50,000, use the promo code **"BOOTSTRAPPED"** to apply for the $5,000 tier instead
- You can request additional credits on a case-by-case basis by explaining your use case in the "Comments" section
- Questions? Email **startups@cloudflare.com**

---

## Program Tiers Overview

| Tier | Credits | Target Stage | Key Requirement |
|---|---|---|---|
| **Bootstrapped** | $5,000 | Stealth / pre-seed | Software startup, founded < 5 years |
| **Growth** | $25,000 | Early traction | Active LinkedIn, funded up to $1M |
| **Seed** | $100,000 | Seed-funded | Funded $1M--$5M, approved VC/accelerator partner |
| **Scale** | $250,000 | Series A+ / AI-focused | Tier 1 VC network, Workers Launchpad, or mission-critical AI app |

All tiers share the same 1-year validity window and the same per-service credit caps.

---

## What the Credits Cover

### Developer Platform (pay-as-you-go services)

| Service | What It Does | Credit Cap |
|---|---|---|
| **Workers** | Serverless compute at the edge | No per-service cap |
| **Pages** | Full-stack web app hosting | No per-service cap |
| **R2** | S3-compatible object storage, zero egress fees | **$10,000 max** |
| **D1** | Serverless SQLite database | No per-service cap |
| **Durable Objects** | Stateful serverless actors | No per-service cap |
| **Workers KV** | Global key-value store | No per-service cap |
| **Queues** | Message queues for async workflows | No per-service cap |
| **Workers AI** | Serverless AI inference (50+ open models) | **$50,000 max** (applies to higher tiers) |
| **Argo Smart Routing** | Optimized traffic routing | No per-service cap |
| **Cache Reserve** | Persistent cache tier | **$10,000 max** (shared with R2) |
| **Images** | Image optimization and resizing | No per-service cap |
| **Cloudflare Calls** | Real-time audio/video | No per-service cap |
| **Stream** | Video encoding and delivery | No per-service cap |

### Enterprise-Level Domains (up to 3 domains)

| Feature | Details |
|---|---|
| **CDN** | Enterprise-grade content delivery |
| **DDoS Protection** | L3/L4/L7 mitigation, no surge pricing |
| **WAF** | Web Application Firewall with managed rulesets |
| **DNS** | Authoritative DNS with DNSSEC |
| **SSL/TLS** | Advanced certificate management |
| **Zero Trust** | Up to 50 seats (Cloudflare for Teams) |
| **Bot Management** | Enterprise bot detection |
| **Rate Limiting** | Advanced rate limiting rules |

---

## Per-Service Credit Caps

Even within the $25,000 total, Cloudflare imposes caps on specific services:

- **R2 + Cache Reserve:** combined maximum of **$10,000** in credits
- **Workers AI:** maximum of **$50,000** (relevant only for $100K and $250K tiers, effectively uncapped at the $25K level)
- All other Developer Platform services share the remaining credit pool with no individual caps

---

## AI Services Available (Workers AI)

Workers AI provides serverless inference on Cloudflare's global network. Key models available:

| Model | Type | Approximate Cost |
|---|---|---|
| **Llama 3.3 70B Instruct (FP8)** | Text generation | $0.29/M input, $2.25/M output tokens |
| **Llama 3.2 11B Vision** | Multimodal (text + image) | $0.05/M input, $0.68/M output tokens |
| **Llama 3.1 8B Instruct (FP8)** | Text generation | $0.05/M input, $0.38/M output tokens |
| **Llama 4 Scout** | Mixture-of-experts, multimodal | Per-model pricing |
| **Mistral Small 3.1** | Text + vision, 128K context | Per-model pricing |
| **Qwen3** | Multilingual reasoning | Per-model pricing |
| **Whisper** | Speech-to-text | Per-model pricing |
| **Stable Diffusion / FLUX** | Image generation | Per-model pricing |

All accounts (including free) get **10,000 Neurons/day free** for Workers AI. Beta models have unlimited free usage until they graduate. With $25K in credits, you can run substantial AI inference workloads.

---

## Cloudflare Free Tier (Always Free, No Credits Needed)

These services are free for all Cloudflare users and do **not** consume your startup credits:

| Service | Free Limit |
|---|---|
| **Workers** | 100,000 requests/day, 10ms CPU per invocation |
| **Pages** | Unlimited sites and bandwidth, 500 builds/month |
| **R2** | 10 GB storage, 1M Class A ops, 10M Class B ops/month |
| **D1** | 5M rows read/day, 100K rows written/day, 5 GB storage |
| **Workers KV** | 100K reads/day, 1K writes/day, 1 GB storage |
| **Durable Objects** | SQLite backend only, 5 GB storage |
| **Workers AI** | 10,000 Neurons/day |
| **DNS** | Unlimited queries, free DNSSEC |
| **DDoS Protection** | Unmetered L3/L4/L7 for all plans |
| **SSL/TLS** | Universal SSL certificate |

Your $25K credits kick in only when you exceed these free limits or use paid-only features.

---

## Workers Launchpad Program

Cloudflare runs an additional **Workers Launchpad** program for startups building on the Developer Platform:

- Hands-on bootcamp sessions with Cloudflare engineers
- Solutions Architect office hours for architecture review
- Introductions to VC partners
- Opportunity to present at **Demo Day**
- Participation can qualify you for the **$250,000 credit tier**

You can apply for Workers Launchpad separately or alongside the startup program.

---

## Coworking Spaces (2026)

Starting in 2026, Cloudflare is opening office locations as **free coworking spaces** for startups in:
- San Francisco
- Austin
- London
- Lisbon

Available to program participants at no cost.

---

## Additional Tips

- **Accelerator affiliation is a bonus, not a requirement** for the $25K tier -- but it strengthens your application. Cloudflare has 250+ approved VC and accelerator partners
- **AI startups no longer need an accelerator referral** -- the program is open to all qualifying AI companies
- **Credits are non-transferable and non-refundable** -- use them or lose them within 12 months
- **Zero egress fees on R2** -- this is a major advantage over AWS S3 or GCP Cloud Storage, and your credits go further because you only pay for storage and operations, not data transfer
- **No auto-billing surprise** -- credits are consumed first; when they run out, you fall back to the free tier or standard pay-as-you-go pricing
- **Request more credits** -- if you are burning through credits quickly, Cloudflare considers case-by-case increases. Explain your use case in the application comments
- **Compare with other startup programs** -- AWS Activate ($5K-$100K), GCP for Startups ($2K-$200K), and Azure for Startups ($1K-$150K) offer similar cloud credits. Cloudflare's advantage is zero egress fees and edge-native serverless compute

---

**Sources:**
- [Cloudflare for Startups](https://www.cloudflare.com/forstartups/)
- [Startup Program Update Blog Post](https://blog.cloudflare.com/expanding-cloudflares-startup-program/)
- [Startup Program Revamp: $250K Credits](https://blog.cloudflare.com/startup-program-250k-credits/)
- [Workers AI Pricing](https://developers.cloudflare.com/workers-ai/platform/pricing/)
- [Workers AI Models](https://developers.cloudflare.com/workers-ai/models/)
- [Workers Pricing](https://developers.cloudflare.com/workers/platform/pricing/)
- [R2 Pricing](https://developers.cloudflare.com/r2/pricing/)
- [D1 Pricing](https://developers.cloudflare.com/d1/platform/pricing/)
- [Developer Platform Pricing](https://www.cloudflare.com/plans/developer-platform/)
- [Cloudflare Startup Enterprise Plan Blog](https://blog.cloudflare.com/the-cloudflare-startup-enterprise-plan-helping-new-startups-bootstrap/)
- [Zero Trust for Startups Reference Architecture](https://developers.cloudflare.com/reference-architecture/design-guides/zero-trust-for-startups/)

## Submission 17: Cloudflare for Startups - Bootstrapped ($5,000)

- Category: Startup Programs
- Source Link: https://www.cloudflare.com/forstartups
- Amount: $5,000
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Bootstrapped and stealth-mode startups can get **$5,000 in Cloudflare credits** valid for one year (or until consumed) by applying with the promo code **BOOTSTRAPPED**. Credits cover pay-as-you-go developer platform services like Workers, Pages, R2, D1, Images, Stream, Queues, and Durable Objects. On top of credits, you get **up to three Enterprise-level domains** with CDN, DDoS protection, DNS, WAF, and Zero Trust included. No VC backing or accelerator membership required — just a software-based startup founded in the last five years.

---

## Registration (Step-by-Step)

1. Go to [cloudflare.com/forstartups](https://www.cloudflare.com/forstartups/)
2. Click **"Apply now"** and create a Cloudflare account (or sign in to an existing one)
3. Fill in the startup application form with your **company name, website, email, and a brief description** of what you are building
4. Enter the promo code **BOOTSTRAPPED** when prompted (this signals you have raised less than $50,000)
5. **Add a valid payment method** — a credit card is required on file. You will not be charged while credits remain, but Cloudflare will bill your card once credits are exhausted or expire
6. Submit the form and check your inbox for a confirmation email that your Startup Plan has been activated

**Important:**
- Acceptance is at **Cloudflare's sole discretion** — not every applicant is approved
- Only **one Startup Plan per applicant** (credits cannot be split across multiple accounts)
- You must be a **new, non-paying Cloudflare user** — existing Enterprise customers are not eligible
- Agencies and service companies are **not eligible**; you must be building a software product
- Your email must match your company domain (generic Gmail/Yahoo addresses may reduce your chances)

---

## What You Get

### Flex Usage Credits ($5,000)

Credits apply to Cloudflare's pay-as-you-go developer platform products:

| Service | What It Does |
|---|---|
| **Workers** | Serverless compute at the edge (JavaScript/TypeScript/Rust/Python) |
| **Pages** | Full-stack web app hosting with Git-based deploys |
| **R2** | S3-compatible object storage with zero egress fees |
| **D1** | Serverless SQLite database |
| **Durable Objects** | Stateful serverless coordination |
| **Queues** | Message queuing between Workers |
| **Workers AI** | Serverless AI inference (50+ models, 200+ edge locations) |
| **Images** | Image optimization, resizing, and delivery |
| **Stream** | Video storage, encoding, and delivery |
| **KV** | Key-value storage at the edge |
| **Hyperdrive** | Connection pooling for existing databases |

### Enterprise-Level Domains (up to 3)

You also get enterprise-grade protection on up to three domains, including:

- **CDN** — global content delivery network
- **DDoS protection** — automatic Layer 3/4/7 mitigation
- **DNS** — authoritative DNS with DNSSEC
- **WAF** — Web Application Firewall with managed rulesets
- **Zero Trust** — identity-based access for internal tools and APIs
- **Argo Smart Routing** — optimized network routing
- **Advanced Certificate Manager** — custom SSL/TLS certificates

---

## Product-Specific Credit Caps

Even within the $5,000 total, certain services have hard sub-limits that apply across all Startup Plan tiers:

| Service | Maximum Credit Usage |
|---|---|
| **R2 + Cache Reserve** | Up to $10,000 (not a constraint at $5K tier) |
| **Workers AI** | Up to $50,000 (not a constraint at $5K tier) |

At the $5,000 tier these caps are not binding, but they matter if you later upgrade to a higher tier ($25K–$250K).

---

## Available AI Models (Workers AI)

Workers AI provides serverless inference on 50+ open-source models. Highlights include:

| Category | Notable Models |
|---|---|
| **Text Generation** | Llama 4 Scout 17B, Llama 3.3 70B, Qwen 3 30B, DeepSeek-R1 32B, Gemma 3 12B, Mistral Small 3.1 24B, GPT-OSS 120B/20B |
| **Code** | Qwen 2.5 Coder 32B |
| **Reasoning** | QwQ-32B, DeepSeek-R1 Distill |
| **Image Generation** | Flux 2 (Klein 9B/4B, Dev), Stable Diffusion XL Lightning, Leonardo Phoenix/Lucid |
| **Speech-to-Text** | Whisper Large v3 Turbo, Deepgram Nova 3 |
| **Text-to-Speech** | Deepgram Aura 2, MeloTTS |
| **Embeddings** | BGE (small/base/large), Qwen 3 Embedding, Google EmbeddingGemma |
| **Vision** | LLaVA 1.5 7B, Llama 3.2 11B Vision |
| **Content Safety** | Llama Guard 3 8B |

Inference is billed at **$0.011 per 1,000 Neurons** (Cloudflare's unified compute unit). Your $5,000 in credits covers this usage.

---

## Cloudflare's Always-Free Baseline

Many Cloudflare services are free regardless of the Startup Plan. These do not consume your credits:

| Service | Free Limit |
|---|---|
| **Workers** | 100,000 requests/day, 10ms CPU/request |
| **Pages** | Unlimited static sites, billed as Workers for Functions |
| **R2** | 10 GB storage, 1M Class A ops, 10M Class B ops/month, zero egress |
| **D1** | 5M rows read/day, 100K rows written/day, 5 GB storage |
| **KV** | 100K reads/day, 1K writes/day, 1 GB storage |
| **CDN** | Unlimited bandwidth (on any plan, including Free) |
| **DNS** | Free authoritative DNS |
| **DDoS** | Unmetered DDoS protection (all plans) |
| **SSL/TLS** | Free Universal SSL certificates |
| **AI Gateway** | 100K logs/month (Free plan) |

Your $5,000 in credits kicks in when you exceed these free limits or use paid-only services (Queues, Stream, Argo, etc.).

---

## What Happens When Credits Expire

1. Credits last **one year** from activation or until fully consumed — whichever comes first
2. Once credits run out, **your credit card on file will be charged** for any ongoing usage above the free tier
3. Services that have always-free tiers (Workers, R2, D1, Pages, CDN, DNS, DDoS) continue at their free limits
4. Paid-only services (Queues, Stream) will stop or start billing your card
5. Enterprise domain protections (WAF, Zero Trust, Argo) revert to free-plan levels unless you subscribe
6. You can request **additional credits** — Cloudflare considers case-by-case extensions at their discretion

---

## Upgrading to Higher Tiers

If your startup grows, you can apply for higher credit tiers:

| Tier | Credits | Requirements |
|---|---|---|
| **$5,000** | Bootstrapped/stealth | Software startup, <5 years old, valid email |
| **$25,000** | Up-and-coming | Above + active LinkedIn, funded up to $1M |
| **$100,000** | Seed-funded | Above + $1M–$5M funding, Cloudflare-approved VC/accelerator |
| **$250,000** | High-growth | Above + Tier 1 VC/accelerator or AI-focused |

**AI startups** get special treatment — you no longer need an accelerator affiliation or employee referral. If you are building in AI with less than $3M in funding, you may qualify for higher tiers directly.

---

## Additional Tips

- **No registrar credits** — domain registration via Cloudflare Registrar is not covered by Startup Plan credits and must be paid separately
- **No enterprise support** — the Startup Plan does not include account management, 24/7 phone support, or SLA guarantees
- **Branding clause** — by accepting, you grant Cloudflare the right to use your company name and logo in their marketing materials
- **Workers Launchpad** — a separate accelerator program offering bootcamp sessions, architect office hours, VC introductions, and Demo Day. Apply separately at [workers.cloudflare.com/launchpad](https://workers.cloudflare.com/launchpad)
- **Non-profit path** — non-profit and civil society organizations have a separate program with free access to developer services
- **Set billing alerts** — since your card is charged automatically after credits expire, monitor your usage in the Cloudflare dashboard to avoid surprise bills
- **Zero egress on R2** — unlike AWS S3 or GCS, R2 has no data transfer charges, making it very cost-effective for storage-heavy workloads even after credits run out

---

**Sources:**
- [Cloudflare for Startups](https://www.cloudflare.com/forstartups/)
- [Startup Program Revamped: Up to $250K in Credits](https://blog.cloudflare.com/startup-program-250k-credits/)
- [Startup Program Update: Empowering Every Stage](https://blog.cloudflare.com/expanding-cloudflares-startup-program/)
- [AI Startups Eligible](https://blog.cloudflare.com/ai-startups-eligible/)
- [Workers Pricing](https://developers.cloudflare.com/workers/platform/pricing/)
- [R2 Pricing](https://developers.cloudflare.com/r2/pricing/)
- [D1 Pricing](https://developers.cloudflare.com/d1/platform/pricing/)
- [Workers AI Pricing](https://developers.cloudflare.com/workers-ai/platform/pricing/)
- [Workers AI Models](https://developers.cloudflare.com/workers-ai/models/)

## Submission 16: ChatGPT Edu - Institutional AI Access

- Category: Education & Student
- Source Link: https://openai.com/chatgpt/education
- Amount: Full ChatGPT access including GPT-5.2, GPT-4.1 mini, o3, o4-mini, Study Mode, Custom GPTs, and agent capabilities (institution-funded, no cost to individual students)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

ChatGPT Edu is OpenAI's university-focused deployment of ChatGPT, giving students and faculty access to advanced AI models with enterprise-grade security and privacy. **If your university has a ChatGPT Edu license, you get full access at no personal cost** — the institution pays. Conversations and data are not used to train OpenAI models. Over a million licenses have been sold to universities worldwide, including massive rollouts at the California State University system (460,000+ users across 23 campuses), Indiana University (120,000 users), University of Oxford, Harvard, USC, Arizona State University, and many others.

---

## How to Get Access

1. **Check with your university IT department** — ask if your institution has a ChatGPT Edu workspace. Many universities announce this through their IT services page or campus-wide emails
2. If your university has ChatGPT Edu, you typically log in via **SSO (Single Sign-On)** using your university credentials (e.g., .edu email)
3. Your workspace admin controls which models and features are enabled — access levels may vary by institution
4. If your university does not have ChatGPT Edu, you can suggest it to your IT department. Universities contact OpenAI's sales team for pricing and deployment
5. **Alternative for US/Canada students:** OpenAI has periodically offered time-limited promotions for individual students (e.g., free ChatGPT Plus for two months in spring 2025 via SheerID verification). Watch for similar offers in the future

---

## What's Included

### AI Models

ChatGPT Edu workspaces provide access to advanced models, with the exact lineup depending on admin configuration:

| Model | Status | Notes |
|---|---|---|
| **GPT-5.2** | Available (Early Access) | Most capable model, admin must enable |
| **GPT-5.1** | Available | Disabled by default, admin can enable |
| **GPT-4.1 mini** | Default | Replaced GPT-4o mini, fast and capable |
| **GPT-4o** | Legacy | Available if admin enables it |
| **o3** | Available | 100 messages/week per user |
| **o4-mini** | Available | 300 messages/day per user |
| **o4-mini-high** | Available | 100 messages/day per user |

### Features

- **Study Mode** — interactive tutoring experience where ChatGPT asks questions to understand your goals and skill level, then guides you to answers collaboratively
- **ChatGPT Agent** — can perform complex online tasks including research across public websites, uploaded files, and connected sources; can fill out forms and edit spreadsheets
- **Custom GPTs** — create and share custom ChatGPT configurations for specific courses, projects, or tasks within the university workspace
- **Shared Projects** — collaborate with teammates by sharing projects with files and instructions that guide ChatGPT responses
- **Advanced Data Analysis** — upload and analyze data files, generate charts, run code
- **Web Browsing / Search** — real-time web search capabilities
- **Multimodal Input** — text, voice, image, and file uploads
- **Canvas** — collaborative writing and coding workspace
- **DALL-E image generation** — available depending on workspace configuration

### Security & Privacy

- Conversations and data are **not used to train OpenAI models**
- Enterprise-grade security with SSO, SCIM provisioning, and group permissions
- Admins can set usage policies, manage access, and monitor adoption
- Data processing compliant with institutional requirements

---

## Limitations

- **Institution-dependent access** — you can only use ChatGPT Edu if your university has purchased a license. There is no way for individual students to buy Edu access directly
- **Usage limits are per-user** — when you hit a model's message limit, that model is temporarily grayed out until the reset time. Limits vary by model (e.g., 100 messages/week for o3)
- **Admin controls what's available** — your university workspace admin decides which models are enabled. You may not have access to all models listed above
- **API access may be restricted** — at some institutions (e.g., Wharton), API access requires faculty sponsorship and is not available to students by default
- **No portability** — your ChatGPT Edu access is tied to your university enrollment. If you graduate or leave, you lose access
- **Contract-dependent duration** — access lasts as long as your university's contract with OpenAI is active
- **Not available at all institutions** — while adoption is growing rapidly, many universities have not yet signed up

---

## Partnered Institutions (Notable Examples)

This is not an exhaustive list. Dozens of universities and entire national education systems have adopted ChatGPT Edu:

- **California State University** — 23 campuses, 460,000+ students, largest single-organization deployment
- **Indiana University** — 120,000 users, second largest rollout
- **University of Oxford** — first UK university to offer ChatGPT Edu to all staff and students
- **Harvard University**
- **University of Southern California (USC)**
- **Arizona State University** — first higher ed institution to partner with OpenAI
- **Columbia University**
- **University of Texas at Austin**
- **Wharton School (University of Pennsylvania)**
- **University of Utah**
- **University of Nebraska Omaha**
- **San Francisco Bay University**

### Education for Countries Initiative

OpenAI's "Edu for Countries" program works with governments to deploy ChatGPT Edu at a national level. The first cohort includes Estonia, Greece, Italy (CRUI consortium), Jordan, Kazakhstan, Slovakia, Trinidad & Tobago, and the United Arab Emirates. Estonia has already deployed to 30,000+ users across public universities and secondary schools.

---

## Additional Tips

- **Check your university email** — many institutions announce ChatGPT Edu rollouts via campus-wide communications. Search your university inbox for "ChatGPT" or "OpenAI"
- **University IT portals** — look for AI tools or software listings on your institution's IT services website
- **Faculty can request it** — if your university does not have ChatGPT Edu, faculty members can reach out to OpenAI's sales team or advocate through their IT governance channels
- **Pricing for institutions** — while not publicly listed, reporting suggests bulk pricing is a few dollars per user per month, significantly less than the $20/month ChatGPT Plus plan
- **Individual student alternatives** — if your university does not offer ChatGPT Edu, consider the free tier of ChatGPT (limited GPT-4o access), or watch for periodic student promotions from OpenAI
- **Data privacy** — unlike the free consumer version of ChatGPT, Edu conversations are not used for model training, making it suitable for research and sensitive academic work
- **Custom GPTs for courses** — ask your professors if they have created course-specific GPTs. Many faculty build custom GPTs for tutoring, language practice, or research assistance

---

**Sources:**
- [ChatGPT Edu — OpenAI](https://openai.com/chatgpt/education/)
- [Introducing ChatGPT Edu — OpenAI Blog](https://openai.com/index/introducing-chatgpt-edu/)
- [ChatGPT Enterprise & Edu Models and Limits — OpenAI Help Center](https://help.openai.com/en/articles/11165333-chatgpt-enterprise-and-edu-models-limits)
- [ChatGPT Enterprise & Edu Release Notes — OpenAI Help Center](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes)
- [OpenAI and the CSU System — OpenAI](https://openai.com/index/openai-and-the-csu-system/)
- [Oxford ChatGPT Edu Announcement — University of Oxford](https://www.ox.ac.uk/news/2025-09-19-oxford-becomes-first-uk-university-offer-chatgpt-edu-all-staff-and-students)
- [Indiana University ChatGPT Edu Rollout — IU News](https://news.iu.edu/live/news/46856-iu-strengthens-national-leadership-in-ai)
- [Education for Countries — OpenAI](https://openai.com/index/edu-for-countries/)
- [Student Discounts for ChatGPT Plus — OpenAI Help Center](https://help.openai.com/en/articles/10968654-student-discounts-for-chatgpt-plus-uscanada)
- [ChatGPT Rate Card (Business, Enterprise/Edu) — OpenAI Help Center](https://help.openai.com/en/articles/11481834-chatgpt-rate-card)

## Submission 15: Cerebras Inference - Free 1M Tokens/Day

- Category: AI API Free Tiers
- Source Link: https://inference.cerebras.ai/
- Amount: 1 million tokens per day
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

Cerebras offers **1 million free tokens per day** through its Inference API, running on proprietary Wafer-Scale Engine (WSE-3) hardware. No credit card required, no waitlist (fully open since June 2025). The API is **OpenAI-compatible** — swap the base URL to `https://api.cerebras.ai/v1` and use your existing OpenAI SDK code. Speeds range from ~1,000 to ~3,000 tokens/second depending on the model, making it one of the fastest inference providers available. Supported models include Llama, Qwen, GPT-OSS, and Z.ai GLM families.

---

## Registration (Step-by-Step)

1. Go to [cloud.cerebras.ai](https://cloud.cerebras.ai)
2. Create an account (email signup)
3. Verify your email address
4. Navigate to **"API Keys"** in the left sidebar
5. Click **"Create API Key"**, give it a name, and copy the key immediately
6. Set it as an environment variable: `export CEREBRAS_API_KEY="your-key-here"`
7. Done — you can start making API calls right away

**Important:**
- No credit card is required for the free tier
- No waitlist or approval process — instant access
- The free tier resets daily (1M tokens per day, not cumulative)
- Unused tokens do not roll over

---

## Available Models

### Production Models

| Model | Model ID | Parameters | Speed | Context (Free) | Status |
|---|---|---|---|---|---|
| **Llama 3.1 8B** | `llama3.1-8b` | 8B | ~2,200 tok/s | 8,192 tokens | Active |
| **Llama 3.3 70B** | `llama-3.3-70b` | 70B | ~2,100 tok/s | 8,192 tokens | Deprecating Feb 16, 2026 |
| **OpenAI GPT-OSS 120B** | `gpt-oss-120b` | 120B | ~3,000 tok/s | 8,192 tokens | Active |
| **Qwen 3 32B** | `qwen-3-32b` | 32B | ~2,600 tok/s | 8,192 tokens | Deprecating Feb 16, 2026 |

### Preview Models (evaluation only, may be discontinued)

| Model | Model ID | Parameters | Speed |
|---|---|---|---|
| **Qwen 3 235B Instruct** | `qwen-3-235b-a22b-instruct-2507` | 235B (22B active MoE) | ~1,400 tok/s |
| **Z.ai GLM 4.7** | `zai-glm-4.7` | 355B | ~1,000 tok/s |

**Note:** Preview models have lower rate limits (especially `zai-glm-4.7` at 10 RPM, 100 RPH, 100 RPD) and are not recommended for production use.

---

## Free Tier Rate Limits

| Limit | Free Tier | Developer Tier (paid) |
|---|---|---|
| **Tokens per Minute** | 60,000 | 1,000,000 |
| **Tokens per Hour** | 1,000,000 | Unlimited |
| **Tokens per Day** | 1,000,000 | Unlimited (pay-as-you-go) |
| **Requests per Minute** | 30 | 1,000 |
| **Requests per Hour** | 900 | Unlimited |
| **Requests per Day** | 14,400 | Unlimited |
| **Context Window** | 8,192 tokens | Up to 128K+ |

Rate limits use a **token bucket** mechanism — capacity replenishes continuously rather than resetting at fixed intervals. Whichever limit (tokens or requests) triggers first will restrict access.

---

## Context Window Limits

The free tier context window is **temporarily limited to 8,192 tokens** across all models. This is a significant constraint for use cases requiring long documents or multi-turn conversations.

On the paid Developer tier, context windows expand substantially — up to **131K tokens** for Qwen 3 235B Instruct and up to **128K** for other models. If you need longer context, the Developer tier starts at just $10.

---

## OpenAI API Compatibility

Cerebras provides an **OpenAI-compatible Chat Completions endpoint**, making migration straightforward. Change two things in your existing code:

1. Set `base_url` to `https://api.cerebras.ai/v1`
2. Use your Cerebras API key instead of your OpenAI key

```python
import openai

client = openai.OpenAI(
    base_url="https://api.cerebras.ai/v1",
    api_key="your-cerebras-api-key"
)

response = client.chat.completions.create(
    model="llama-3.3-70b",
    messages=[{"role": "user", "content": "Hello!"}]
)
```

**Unsupported OpenAI parameters:**
- `frequency_penalty`, `presence_penalty`, `logit_bias`
- Streaming + JSON mode on reasoning models (streaming works fine with `gpt-oss-120b` and non-reasoning models)
- Text Completions endpoint (only Chat Completions is supported)

Cerebras also provides native SDKs for **Python** (`pip install cerebras_cloud_sdk`) and **Node.js** (`npm install @cerebras/cerebras_cloud_sdk`).

---

## Paid Tiers

| Tier | Cost | Key Differences |
|---|---|---|
| **Free** | $0 | 1M tokens/day, 8K context, community support |
| **Developer** | From $10 (pay-as-you-go) | 10x higher rate limits, up to 128K+ context, no daily cap |
| **Enterprise** | Custom pricing | Dedicated queue priority, custom model weights, fine-tuning, SLA support |

The Developer tier has no contracts — deposit $10 to your account and pay per token consumed. No auto-billing traps on the free tier.

---

## Alternative Access Methods

You can also access Cerebras-powered inference through:
- **[OpenRouter](https://openrouter.ai/provider/cerebras)** — unified API supporting multiple providers
- **[Hugging Face Inference](https://huggingface.co/docs/inference-providers/en/providers/cerebras)** — access Cerebras models from the HF Hub
- **AWS Marketplace** — Cerebras Fast Inference Cloud available as a marketplace product

---

## Additional Tips

- **1M tokens/day is enough** for serious prototyping, small internal tools, or early-stage pilots — but the 8K context window on free tier is the real bottleneck for many use cases
- **Model deprecation** — Llama 3.3 70B and Qwen 3 32B are scheduled for deprecation on **February 16, 2026**. Plan migrations to newer models
- **`gpt-oss-120b` system role behavior** — the `system` role maps to developer-level instructions with stronger influence than standard OpenAI, so identical prompts may produce different results
- **Cerebras Code** — a separate product (VS Code extension) with Pro ($50/mo, 24M tokens/day) and Max ($200/mo, 120M tokens/day) plans for coding assistance
- **No region restrictions** — the API is globally available, with data centers across North America and Europe
- **Speed advantage** — Cerebras benchmarked Llama 4 Maverick at 2,522 tok/s vs. NVIDIA Blackwell at 1,038 tok/s for the same model

---

**Sources:**
- [Cerebras Inference](https://www.cerebras.ai/inference)
- [Cerebras Pricing](https://www.cerebras.ai/pricing)
- [Cerebras Supported Models](https://inference-docs.cerebras.ai/models/overview)
- [Cerebras Rate Limits](https://inference-docs.cerebras.ai/support/rate-limits)
- [Cerebras Quickstart](https://inference-docs.cerebras.ai/quickstart)
- [Cerebras OpenAI Compatibility](https://inference-docs.cerebras.ai/resources/openai)
- [Cerebras Inference Free Tier Analysis (Adam Holter)](https://adam.holter.com/cerebras-opens-a-free-1m-tokens-per-day-inference-tier-and-claims-20x-faster-than-nvidia-real-benchmarks-model-limits-and-why-ui2-matters/)
- [CloudCredits.io - Cerebras Free Tier](https://cloudcredits.io/providers/cerebras/programs/cerebras-free-llm-credits)
- [Cerebras Blog: Pay Per Token](https://www.cerebras.ai/blog/cerebras-inference-now-available-via-pay-per-token)

## Submission 14: Azure for Students - $100/Year (No Credit Card)

- Category: Education & Student
- Source Link: https://azure.microsoft.com/en-us/free/students
- Amount: $100 per year (renewable annually while enrolled)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

Verified students at accredited institutions get **$100 in Azure credits for 12 months** with **no credit card required**. The offer is renewable annually as long as you remain enrolled. You also get access to **25+ free Azure services** (compute, storage, databases, AI/ML) and **software downloads** through the Azure Education Hub. Microsoft markets access to Azure OpenAI models including GPT-5 series, GPT-4.1, and o3/o4-mini reasoning models, but in practice **student subscriptions have severely limited quotas** (as low as 1,000 tokens per minute across all models) and many students report deployment errors.

---

## Registration (Step-by-Step)

1. Go to [azure.microsoft.com/free/students](https://azure.microsoft.com/en-us/free/students)
2. Click **"Start free"** or **"Activate now"**
3. Sign in with a **personal Microsoft account** (e.g. @outlook.com, @hotmail.com). Do **not** use your .edu or institutional email to sign in -- use it only for verification in a later step. Create a Microsoft account if you do not have one
4. **Phone verification** -- enter your phone number and verify via SMS or voice call
5. **Student identity verification** -- enter your first name, last name, country, date of birth, school name (selected from a dropdown), and your **institutional email address** (typically ending in .edu or .ac)
6. **Email verification** -- Microsoft sends a verification link to your institutional email. Click the link to confirm your student status
7. Done -- your subscription is created with $100 in credits, no credit card needed

**Important:**
- One subscription per eligible student -- non-transferable, cannot be combined with other offers
- The 12-month clock starts at activation, not at first usage
- Unused credits do **not** roll over to the next year
- If your institution is not in Microsoft's dropdown list, you may be asked to upload a photo of your student ID card instead

---

## Eligibility Requirements

- **Age:** 18 or older
- **Institution:** Accredited, degree-granting two-year or four-year educational institution
- **Enrollment:** Full-time student
- **Verification:** Must have an institutional email address (.edu, .ac, or equivalent)
- **Excluded:** MOOC participants, for-profit professional training programs, audit-only enrollment

### Alternative Verification (No .edu Email)

If your school does not provide a .edu email:
- **GitHub Student Developer Pack** -- if approved through GitHub, Microsoft may accept your GitHub Student verification as proof of student status
- **Manual verification** -- contact Azure Support and upload enrollment proof or student ID. Response time is typically 2-3 business days
- **Activation code** -- some institutions distribute Azure activation codes directly; check with your university IT department

---

## Azure OpenAI Access (Reality Check)

Microsoft's marketing page advertises access to GPT-4 series, o1 series, and 1,700+ foundation models through Azure AI Foundry. The reality for student subscriptions is more limited:

### What Actually Works
- You **can** create an Azure OpenAI resource and deploy models
- GPT-4o-mini and GPT-4.1-nano are the most reliably deployable models on student subscriptions
- Some students successfully deploy GPT-4o with very low quotas

### Common Limitations
- **Maximum quota of ~1,000 tokens per minute (TPM)** across all models -- far below standard subscription quotas
- **"Insufficient quota" errors** are extremely common when deploying GPT-5 series, GPT-4.1, or higher-end models
- **Region restrictions** -- student subscriptions have Azure Policy restrictions that may block deployment in regions where Azure OpenAI is available (e.g. East US 2, Sweden Central)
- The Assistant Playground and Agent Playground may be effectively unusable due to quota limits

### Workarounds
1. **Try East US, South Central US, or West Europe** -- these regions tend to have better support for student OpenAI deployments
2. **Request a quota increase** -- Azure Portal > Help + Support > New Support Request > Quota Increase. Provide a justification (e.g. academic project). Approval is not guaranteed and takes 1-2 days
3. **Deploy GPT-4o-mini or GPT-4.1-nano first** -- they require less quota and are more likely to succeed
4. **Check if your university provides institutional Azure OpenAI access** -- many universities have separate agreements with higher quotas
5. **Upgrade to Pay-As-You-Go** -- removes region and quota restrictions but requires a credit card

---

## AI-Related Free Services

### Always Free (ongoing, within monthly limits)

| Service | Free Limit |
|---|---|
| **Azure AI Search** | 50 MB storage, 10,000 documents, 3 indexes |
| **Language Service** (sentiment, key phrases, NER) | 5,000 text records/month |
| **Content Safety** | 5,000 text records or images/month |
| **Speech to Text** | 5 audio hours/month |
| **Text to Speech** | 500,000 neural characters/month |
| **Translator** | 2 million characters/month |
| **Azure Machine Learning** | Free tier (R and Python model development) |

### 12 Months Free

| Service | Free Limit |
|---|---|
| **Azure Vision** (Computer Vision) | 5,000 transactions/month (S1, S2, S3 tiers) |
| **Document Intelligence** | 500 pages/month (S0 tier) |
| **AI Custom Vision** | 10,000 predictions + 1 training hour + 2 projects |
| **Face API** | 30,000 transactions/month (S0 tier) |

### Foundry Models Catalog

Azure AI Foundry provides access to 1,700+ foundation models from Microsoft, OpenAI, Hugging Face, Meta, and Cohere. However, deploying these models consumes your $100 credits and may be limited by student subscription quotas.

---

## Non-AI Free Services (Highlights)

### 12 Months Free

| Service | Free Limit |
|---|---|
| **Linux VMs** | 750 hours B1s, B2pts v2, B2ats v2 |
| **Windows VMs** | 750 hours B1s, B2pts v2, B2ats v2 |
| **PostgreSQL Flexible Server** | 750 hours B1MS, 32 GB storage |
| **MySQL Flexible Server** | 750 hours B1MS, 32 GB storage |
| **Container Registry** | 1 Standard registry, 100 GB storage |

### Always Free

| Service | Free Limit |
|---|---|
| **App Service** | 10 web/API apps, 1 GB storage |
| **Azure Functions** | 1 million requests/month |
| **Container Apps** | 180,000 vCPU seconds, 360,000 GiB seconds |
| **Azure Cosmos DB** | 1,000 RU/s, 25 GB storage |
| **Azure SQL Database** | 100,000 vCore seconds, 32 GB storage |
| **Azure DevOps** | 5 users, unlimited private Git repos |
| **Azure Kubernetes Service** | Cluster management free (node costs apply from credits) |

---

## Credit Restrictions

The $100 credits **cannot** be used for:
- Azure support plans
- Azure DevOps paid tiers
- Visual Studio subscriptions or App Center
- Azure ExpressRoute
- Third-party/branded products from Azure Marketplace
- Products in sovereign cloud regions (US Government, Azure China, Azure Germany)

---

## Renewal Process

1. Microsoft sends a reminder email before your 12-month anniversary
2. Go to [azure.microsoft.com/free/students](https://azure.microsoft.com/en-us/free/students) and sign up again
3. Re-verify your student status using your institutional email
4. If verified, you receive a fresh $100 in credits for another 12 months
5. Unused credits from the previous year do **not** carry over

**If you are no longer a student:**
- You cannot renew
- You have 90 days to upgrade to a Pay-As-You-Go subscription
- After 90 days, deployed resources are decommissioned

---

## What Happens When Credits Run Out

- Your subscription is **disabled** -- all running services are stopped
- You are **not charged** -- there is no credit card on file
- **Option A:** Renew if still a student and the 12-month period has ended
- **Option B:** Upgrade to Pay-As-You-Go by contacting Azure Support (requires adding a credit card)
- **Option C:** Do nothing -- subscription is canceled, resources are decommissioned
- Monitor remaining credit at [microsoftazuresponsorships.com](https://www.microsoftazuresponsorships.com/)

---

## Software Downloads (Education Hub)

Azure for Students includes free software downloads through the Azure Education Hub:
- Development tools (Visual Studio, etc.)
- Software intended for education, non-commercial research, and application development/testing
- Access via the Education Hub section in the Azure Portal

---

## Additional Tips

- **Azure Free Account ($200 credits)** is a separate program for everyone (not just students) but **requires a credit card**. If you have both options, the student program is better for long-term use since it renews annually
- **Microsoft for Startups Founders Hub** offers up to $5,000-$150,000 in Azure credits if you are building a startup -- this can be combined with student status
- **Set budget alerts** in the Azure Portal to avoid accidentally exhausting your $100 credits on expensive services (GPU VMs can burn through credits in hours)
- **Deallocate VMs when not in use** -- a stopped VM without deallocation still incurs charges against your credits
- **Use the Azure Pricing Calculator** before deploying resources to estimate how long your credits will last
- **Common verification issues** -- if you get "Unable to confirm your University ID" or "No permissions" errors, try clearing browser cache, using incognito mode, or contacting Azure Support with your student ID

---

**Sources:**
- [Azure for Students Signup](https://azure.microsoft.com/en-us/free/students)
- [Azure for Students Offer Details](https://azure.microsoft.com/en-us/pricing/offers/ms-azr-0170p/)
- [About Azure for Students Program](https://learn.microsoft.com/en-us/azure/education-hub/about-azure-for-students)
- [Azure for Students Program Details](https://learn.microsoft.com/en-us/azure/education-hub/azure-dev-tools-teaching/azure-students-program)
- [Azure Education Hub FAQ](https://learn.microsoft.com/en-us/azure/education-hub/azure-dev-tools-teaching/program-faq)
- [Explore Free Azure Services](https://azure.microsoft.com/en-us/pricing/free-services)
- [Azure OpenAI with Azure for Students (Q&A)](https://learn.microsoft.com/en-us/answers/questions/2183197/azure-openai-with-azure-for-students)
- [Azure AI Foundry Quota Issues - Student Account](https://learn.microsoft.com/en-us/answers/questions/2224916/azure-ai-foundry-quota-insufficient-issues-student)
- [Azure OpenAI Quotas and Limits](https://learn.microsoft.com/en-us/azure/ai-services/openai/quotas-limits)

## Submission 13: AWS Free Tier - Up to $200 Credits (2025+)

- Category: Cloud Providers
- Source Link: https://aws.amazon.com/free
- Amount: Up to $200 ($100 signup + $100 onboarding tasks)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

New AWS customers (accounts created after July 15, 2025) receive **up to $200 in credits**: $100 on signup plus $100 earned by completing 5 onboarding tasks ($20 each). You choose between a **Free plan** (zero-charge guarantee, 6 months) or a **Paid plan** (pay-as-you-go after credits). Credit card required but you will **not be charged on the Free plan** even if credits run out -- AWS simply closes the account. Includes access to 200+ services including Amazon Bedrock AI models.

---

## Registration (Step-by-Step)

1. Go to [aws.amazon.com/free](https://aws.amazon.com/free) and click **"Create an AWS Account"**
2. Enter your **email address** and choose an AWS account name
3. **Choose your plan**: Free or Paid (you can upgrade later, but cannot downgrade)
4. Enter **billing information** -- a valid credit or debit card is required. AWS places a temporary **$1 USD hold** (released in 3-5 days) to verify your identity
5. **Identity verification** -- receive a code via SMS or voice call
6. Account activation takes **30-60 minutes** (email confirmation sent)
7. Log into the **AWS Management Console** and find the **"Explore AWS"** widget to start onboarding tasks

**Important:**
- The card must be **new to AWS** -- it cannot have been used for another AWS account
- Some banks may ask for additional verification for online payments
- Prepaid cards may not work
- One free tier per person/account

---

## Credit Structure: $100 + $100

### Instant Credit: $100
Applied immediately to your account upon signup, regardless of which plan you choose.

### Onboarding Tasks: $100 ($20 each)

Complete these 5 guided activities in the AWS Console to earn $20 per task:

| Task | What You Do | Credit |
|---|---|---|
| **Amazon EC2** | Launch an EC2 instance and terminate it | $20 |
| **Amazon RDS** | Configure and launch an RDS database | $20 |
| **AWS Lambda** | Build a simple web app with a Lambda function URL | $20 |
| **Amazon Bedrock** | Submit a prompt in the Bedrock text playground | $20 |
| **AWS Budgets** | Set a budget alert for cost monitoring | $20 |

Track progress in the **"Explore AWS"** widget in the Console. Each task takes 5-15 minutes.

**Credit expiration:** All credits (both $100 batches) expire **12 months** from account creation, regardless of plan.

---

## Free Plan vs. Paid Plan

| | Free Plan | Paid Plan |
|---|---|---|
| **Cost guarantee** | Zero charges, ever | Pay-as-you-go after credits |
| **Duration** | 6 months or until credits exhausted | Indefinite |
| **Services** | 200+ services (some enterprise services restricted) | All 200+ services |
| **What happens at expiry** | Account closed, 90-day data retention | Continues with standard billing |
| **Credits** | $200 max, expire at 12 months | $200 max, expire at 12 months |
| **Can upgrade?** | Yes, anytime | N/A |
| **Can downgrade?** | N/A | No, cannot go back to Free |

**Recommendation for builders:** Start with the **Free plan**. You can upgrade anytime and keep your remaining credits (valid up to 12 months from signup).

---

## What Changed After July 15, 2025

- **Old model (pre-July 15):** 12-month free trial with per-service usage limits (e.g., 750 hrs/month EC2 t2.micro). No upfront credits. Pay-as-you-go if you exceed limits.
- **New model (post-July 15):** Credit-based system ($200 total). Free plan with zero-charge guarantee. Account auto-closes instead of billing you.
- **Legacy accounts** created before July 15, 2025 remain on the old 12-month trial model -- they do not get the $200 credits.
- The 12-month free trial for new accounts has been **discontinued**.

---

## AI Services

### Amazon Bedrock (Credits-Based)

Bedrock is **not** in the Always Free tier. Usage is billed per token and deducted from your $200 credits. Available models include:

- **Anthropic Claude** (Claude Opus 4.5, Claude Sonnet 4.5, Claude Haiku 4.5, Claude Sonnet 4, Claude 3.5 Haiku)
- **Meta Llama** (Llama 4 Maverick, Llama 4 Scout, Llama 3.3 70B, Llama 3.2, Llama 3.1)
- **Mistral AI** (Mistral Large 3, Pixtral Large, Magistral Small, Ministral, Voxtral)
- **Amazon Nova** (Nova Premier, Nova Pro, Nova Lite, Nova Micro, Nova Sonic, Nova Canvas, Nova Reel)
- **Amazon Titan** (Text, Embeddings, Image)
- **DeepSeek** (DeepSeek-R1, DeepSeek-V3.1)
- **Qwen** (Qwen3 235B, Qwen3 Coder, Qwen3 VL)
- **Cohere** (Command R+, Command R, Embed v4)
- **Stability AI** (Stable Diffusion 3.5 Large, Stable Image Ultra/Core)
- **OpenAI** (GPT-OSS 120B, GPT-OSS 20B)

Best model variety is in **us-east-1** (N. Virginia) and **us-west-2** (Oregon).

### Other AI/ML Services

| Service | Free Tier Type | Limits | Duration |
|---|---|---|---|
| **Amazon SageMaker** | 12-month (legacy accounts) / Credits (new) | 250 hrs t2.micro/month for training | 12 months |
| **SageMaker Studio Lab** | Always Free | Jupyter ML environment (CPU/GPU), no AWS account needed | Indefinite |
| **Amazon Rekognition Image** | Credits-based (new) | ~1,000 images/month equivalent | Until credits run out |
| **Amazon Rekognition Video** | Credits-based (new) | ~60 min video analysis/month equivalent | Until credits run out |
| **Amazon Comprehend** | Credits-based (new) | 50,000 units of text per API equivalent | Until credits run out |
| **Amazon Polly** | Credits-based (new) | ~5M characters/month equivalent | Until credits run out |
| **Amazon Transcribe** | Credits-based (new) | ~60 min/month equivalent | Until credits run out |
| **Amazon Translate** | Credits-based (new) | ~2M characters/month equivalent | Until credits run out |

**Note:** For new accounts (post-July 2025), the old per-service 12-month limits are replaced by the unified $200 credit pool. Your credits cover any eligible service usage.

---

## Always Free Services (Highlights)

These have **permanent monthly limits** that don't consume credits and persist for the life of the account:

| Service | Monthly Free Limit |
|---|---|
| **AWS Lambda** | 1M requests, 400K GB-seconds |
| **Amazon DynamoDB** | 25 GB storage, 25 RCU/WCU |
| **Amazon S3** | 5 GB Standard storage |
| **Amazon CloudFront** | 1 TB data out, 10M requests |
| **Amazon SNS** | 1M publishes |
| **Amazon SQS** | 1M requests |
| **Amazon CloudWatch** | 10 custom metrics, 10 alarms |
| **AWS Step Functions** | 4,000 state transitions |
| **Amazon API Gateway** | 1M REST API calls |
| **AWS CodeBuild** | 100 build-minutes |
| **AWS CodePipeline** | 1 active pipeline |

---

## Billing Gotchas and Warnings

1. **Organizations trap:** If your account joins an AWS Organization or sets up AWS Control Tower, your Free Tier credits **expire immediately** and your account is auto-upgraded to Paid. Do not join an Organization during your free period.
2. **No downgrade:** Once you upgrade from Free to Paid, you **cannot go back**. Choose carefully.
3. **Free plan account closure:** When your 6-month free plan expires, AWS **closes your account**. You have 90 days to upgrade and recover data. After 90 days, everything is **permanently deleted**.
4. **Paid plan auto-billing:** On the Paid plan, once credits are exhausted, usage transitions **seamlessly to standard billing** with no warning gate. Set up Budgets immediately.
5. **Restricted services on Free plan:** Savings Plans, Reserved Instances, and certain AWS Marketplace offers are blocked on the Free plan to prevent accidental credit depletion.
6. **Credit notifications:** AWS sends alerts at 50%, 25%, and 10% credit remaining, plus at 15, 7, and 2 days before the 6-month expiry.
7. **Aggregated usage in Organizations:** If somehow in an Organization, free tier usage is aggregated across all member accounts.

---

## Tips for Indie Builders and Startups

- **Do the Bedrock onboarding task first** -- it takes 5 minutes (just send a prompt in the playground) and earns you $20 in credits while introducing you to foundation models.
- **Do the Budgets task immediately** -- it teaches you to set up cost alerts, which protects you from surprise bills if you ever upgrade. Also earns $20.
- **Complete all 5 tasks on day one** -- it takes under an hour and doubles your credits from $100 to $200.
- **Start with the Free plan** -- zero risk. Upgrade to Paid only when you're ready for production.
- **Use us-east-1 or us-west-2** for AI workloads -- best model availability on Bedrock.
- **SageMaker Studio Lab is free forever** -- use it for ML experimentation without touching your credits.
- **Check AWS Activate** if you're a startup -- up to $100K in credits (up to $300K for AI startups in 2026).
- **Check AWS Educate** if you're a student -- additional $200 in credits for learning.
- **Monitor credits** in the AWS Billing and Cost Management console regularly.
- **Terminate resources you're not using** -- even on the Free plan, idle resources consume credits faster than you'd expect.

---

**Sources:**
- [AWS Free Tier](https://aws.amazon.com/free/)
- [AWS Blog: Free Tier Update - Up to $200 in Credits](https://aws.amazon.com/blogs/aws/aws-free-tier-update-new-customers-can-get-started-and-explore-aws-with-up-to-200-in-credits/)
- [AWS Free Tier Announcement (July 2025)](https://aws.amazon.com/about-aws/whats-new/2025/07/aws-free-tier-credits-month-free-plan/)
- [AWS Free Tier FAQs](https://aws.amazon.com/free/free-tier-faqs/)
- [AWS Registration FAQs](https://aws.amazon.com/free/registration-faqs/)
- [AWS Billing: Free Tier Plans](https://docs.aws.amazon.com/awsaccountbilling/latest/aboutv2/free-tier-plans.html)
- [AWS Billing: Earning Additional Credits](https://docs.aws.amazon.com/awsaccountbilling/latest/aboutv2/free-tier-plans-activities.html)
- [Amazon Bedrock Pricing](https://aws.amazon.com/bedrock/pricing/)
- [AWS Free Tier 2025 Overview (DEV Community)](https://dev.to/aws-builders/aws-free-tier-2025-how-to-get-200-in-credits-and-what-changed-3cj0)
- [AWS Free Tier GitHub Reference (costgoat)](https://github.com/costgoat/aws-free-tier)

## Submission 12: AWS Educate - Free Cloud Training & Credits for Students

- Category: Education & Student
- Source Link: https://aws.amazon.com/education/awseducate
- Amount: Up to $100 in promotional credits (students) or $200 (educators) + free hands-on labs and training
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

AWS Educate gives students and educators **free access to cloud computing training**, hands-on labs, digital badges, and AWS promotional credits -- all **without a credit card**. Learners as young as 13 can register with just an email address. The program focuses on guided career pathways, sandbox-style labs (Amazon S3, EC2, RDS, VPC), and a job board for learners 18+. Credit amounts range from **$25 to $100** depending on institutional affiliation, though the program has shifted toward sandbox labs rather than open-ended AWS Console access.

---

## Registration (Step-by-Step)

1. Go to [aws.amazon.com/education/awseducate](https://aws.amazon.com/education/awseducate/) and click **"Register Now"**
2. Fill out the registration form: **name, email, date of birth, location, and language preferences**
3. In the program association dropdown, select **"AWS Educate"** (optionally also "AWS Academy")
4. Click **"Create Account"**
5. Check your email for a **verification link** from AWS and click **"Verify my email"**
6. Wait for approval -- typically **within 30 minutes**, you receive an approval email with a password setup link
7. Set your password and log into the AWS Educate portal

**Important:**
- **No credit card required** -- the entire program is free
- Use an **institutional/school email** (.edu or equivalent) for best results; personal emails also work but may receive fewer credits
- Learners **age 13+** can register; the **job board** requires age 18+
- Approval is typically fast (under an hour) but can vary

---

## Credit Amounts

The amount of AWS promotional credits depends on your role and whether your institution is an AWS Educate member:

| Role | Institution Member | Non-Member |
|---|---|---|
| **Student** | $100 | $25-$35 |
| **Educator** | $200 | $75 |

**Credit details:**
- Credits are valid for **up to 12 months**
- Credits are applied to an **AWS Educate Starter Account**, which is separate from a regular AWS account
- Credits **cannot** be used for: Amazon Mechanical Turk, AWS Managed Services, AWS Marketplace, Route 53 domain registration, cryptocurrency mining, Reserved Instances, or Savings Plans
- Credits are for **learning and coursework only** -- not designed for production workloads

---

## Starter Account vs. Regular AWS Account

The AWS Educate Starter Account is **not** a regular AWS account. Key differences:

| Feature | Starter Account | Regular AWS Account |
|---|---|---|
| **Credit card** | Not required | Required |
| **Region** | us-east-1 only | All regions |
| **Services** | ~75% of AWS services | Full access |
| **IAM** | Not available | Full access |
| **Billing dashboard** | Not available | Full access |
| **EC2 instance types** | Limited (t2.nano through t2.2xlarge, m4/m5, c4/c5) | All types |
| **Instance limit** | 20 concurrent instances | Standard limits |
| **Session timeout** | 3 hours of activity | No timeout |
| **Post-graduation** | Account expires; must renew annually | Portable forever |
| **Risk of charges** | Zero -- impossible to be billed | Yes, if Free Tier exceeded |

**Bottom line:** The Starter Account is a safe sandbox for learning with zero financial risk, but it is significantly more restricted than a full AWS account.

---

## Career Pathways & Learning Content

AWS Educate offers **25+ Cloud Career Pathways** -- self-paced learning modules with a minimum of 30 hours of content each. Pathways span four job families:

| Job Family | Example Pathways |
|---|---|
| **Cloud Architect** | Cloud infrastructure design, Well-Architected Framework |
| **Software Developer** | Application development, serverless, containers |
| **Operations & Support** | Monitoring, deployment, systems administration |
| **Data & Analytics** | Big data, data analytics, machine learning |

Additional specialized pathways include **Cybersecurity** and **Machine Learning**.

### Hands-On Labs

Free sandbox labs covering core AWS services:

- **Amazon S3** -- build cloud storage, host static websites
- **Amazon EC2** -- launch and manage compute instances
- **Amazon RDS** -- set up and connect databases
- **Amazon VPC** -- configure networking and security
- **AWS Cost Estimation** -- learn cloud operations and budgeting

Labs run in **temporary sandboxed AWS environments** so you cannot accidentally rack up charges.

---

## Digital Badges

Complete pathways and courses to earn **shareable digital badges** via Credly:

- **Cloud Literacy Badge** -- foundational cloud concepts
- **IoT Badge** -- Internet of Things specialization
- **Alexa Badge** -- voice interface development
- **Start-up Badge** -- cloud for startups
- **Gaming Badge** -- game development on AWS

Badges are unlocked after completing **50% of any career pathway**, followed by knowledge checks, projects, and a final exam. Badges can be displayed on LinkedIn, portfolios, and resumes.

---

## Job Board (18+ Only)

Learners aged 18 and older gain access to the **AWS Educate Job Board**, which features thousands of cloud-related jobs and internships from organizations worldwide. The job board is integrated with the career pathways, so completing training directly surfaces relevant opportunities.

---

## Who This Is For (and Who It's Not For)

**Good fit:**
- Students currently enrolled in a university, college, or qualifying high school
- Educators teaching cloud computing or related subjects
- Beginners who want zero-risk exposure to AWS with no credit card
- Anyone building a portfolio of cloud skills and badges

**Not ideal for:**
- Self-taught developers or bootcamp graduates without institutional email -- you may only get read-only learning materials without sandbox access
- Anyone needing production-ready infrastructure or full AWS service access
- Indie builders who need Bedrock, SageMaker, or advanced AI services -- these are **not available** in Starter Accounts
- Anyone who wants to keep their AWS environment after graduation

---

## Program Changes (2023-2025)

- **February 2023:** AWS Educate stopped offering grants to institutions and educators directly. Educator tools migrated to **AWS Academy**
- **2024-2025:** The program has been shifting from the traditional Starter Account (managed by Vocareum/Qwiklabs) toward a **sandbox lab-based model** with guided, time-boxed exercises
- **July 2025:** AWS overhauled its general Free Tier (separate from Educate) to a credit-based model with up to $200 for new accounts -- this is a separate program that does require a credit card

---

## Additional Tips

- **Combine with AWS Free Tier** -- if you also create a regular AWS account (credit card required), you get up to $200 in credits on top of AWS Educate. The two programs are separate and stack
- **AWS Academy** is the institutional counterpart -- if your school participates, you may get richer lab environments through Academy rather than Educate
- **AWS Cloud Institute** is a newer program offering a full virtual training program for cloud careers (paid, but scholarships available)
- **SageMaker Studio Lab** is always free and does **not** require an AWS account -- great for ML experimentation alongside Educate
- **Set up a regular AWS account alongside Educate** if you need full service access; use Educate for structured learning and the regular account for building
- **Complete pathways before graduation** -- your Starter Account access expires when you leave your institution
- **LocalStack** is a free alternative for local AWS development and testing if the Starter Account restrictions are too limiting

---

**Sources:**
- [AWS Educate Official Page](https://aws.amazon.com/education/awseducate/)
- [AWS Educate Career Pathways & Badges (AWS Blog)](https://aws.amazon.com/blogs/publicsector/aws-educate-gives-students-access-to-cloud-career-pathways-aws-badges-and-the-aws-educate-job-board/)
- [AWS Educate Starter Account Services (PDF)](https://awseducate-starter-account-services.s3.amazonaws.com/AWS_Educate_Starter_Account_Services_Supported_v1.pdf)
- [AWS Educate Starter Account (GeeksforGeeks)](https://www.geeksforgeeks.org/cloud-computing/aws-educate-starter-account/)
- [How to Get AWS Educate Account (DEV Community)](https://dev.to/muhammad_abdullah_0f40cb1/how-to-get-an-aws-educate-account-no-credit-card-required-a-step-by-step-guide-for-students-3f6j)
- [What AWS Educate Really Offers (Tech Reader)](https://www.tech-reader.blog/2025/04/insight-what-aws-educate-really-offers.html)
- [How to Get Free AWS Credits (Spendbase)](https://www.spendbase.co/blog/spend-management/how-to-get-free-aws-credits/)
- [Free AWS Credits (Cloudvisor)](https://cloudvisor.co/how-to-get-free-aws-credits/)
- [AWS Educate Training Pathways (AWS Blog)](https://aws.amazon.com/blogs/publicsector/a-commitment-to-training-learning-pathways-to-real-world-jobs-with-aws-educate/)

## Submission 11: AWS Activate for Y Combinator - $500,000

- Category: Startup Programs
- Source Link: https://aws.amazon.com/startups/credits
- Amount: $500,000 (~$200K cloud + ~$300K Trainium/Inferentia AI silicon)
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Y Combinator batch companies receive one of the most generous cloud credit packages available through any accelerator: **up to $500,000 in AWS credits**, split between general cloud credits and AI-specific compute credits for AWS Trainium and Inferentia silicon. Credits are redeemable across **200+ AWS services** including Amazon Bedrock (with third-party models from Anthropic, Meta, Mistral AI, and others), Amazon EC2, Amazon SageMaker, and reserved GPU capacity. This package dwarfs the standard AWS Activate Portfolio tier ($100K–$200K) and positions AWS as the default cloud for YC-backed AI startups.

---

## Credit Breakdown

| Component | Amount | What It Covers |
|---|---|---|
| **General Cloud Credits** | ~$200,000 | Any eligible AWS service: compute (EC2, Lambda), storage (S3), databases (RDS, DynamoDB), networking, analytics, and more |
| **AI Silicon Credits** | ~$300,000 | AWS Trainium (training) and Inferentia (inference) chip instances, Amazon Bedrock model usage, and reserved NVIDIA H100 GPU capacity |
| **Total** | **$500,000** | Combined across all eligible services |

**Note:** The exact breakdown may vary by YC batch and is negotiated between AWS and Y Combinator. The $500K figure was confirmed for the W24 (Winter 2024) cohort. Future cohorts may receive similar or adjusted packages.

---

## Eligibility

- **Active Y Combinator batch company** — you must be accepted into a current or recent YC batch (W24 onward has confirmed $500K)
- Company must be **less than 10 years old**
- Must be **pre-Series B** (or self-funded)
- Must not have previously received AWS Activate Portfolio credits of equal or greater value
- Must have a **valid AWS account** with a credit card on file
- Must have a **functioning company website** with a matching business email domain

This package is **not available** to YC alumni from older batches, YC Startup School participants, or companies that only applied to YC but were not accepted. Those groups can still apply for the standard AWS Activate Portfolio tier (up to $200K).

---

## How to Redeem (Step-by-Step)

1. **Get accepted into Y Combinator** — the credit package is distributed through YC's partnership with AWS, not through the public Activate application
2. **Receive your Organization ID (Org ID)** from YC — Y Combinator is an AWS Activate Provider and will share the specific Org ID and any promo codes during onboarding
3. **Create or sign in to your AWS account** at [aws.amazon.com](https://aws.amazon.com) — you need a credit card on file (no charge unless you exceed credits)
4. **Go to [AWS Activate](https://aws.amazon.com/activate/)** and click **Apply Now**
5. **Create an AWS Builder ID** using your business email (must match your startup’s domain, e.g., founder@yourstartup.com — personal Gmail addresses are a common reason for rejection)
6. **Select the Portfolio Package** and enter the Org ID provided by Y Combinator
7. **Complete your startup profile** — company name, website, description, stage, AWS Account ID
8. **Submit the application** — processing typically takes **7–10 business days**
9. **Credits appear automatically** in your AWS Billing Console under **Billing > Credits** once approved

**Important caveats:**
- Use a **business email** that matches your company domain. Personal emails (Gmail, Yahoo) are the #1 cause of rejection
- You cannot choose which services consume credits first — AWS applies credits automatically to the largest charges first
- Credits that expire soonest are applied first
- Apply as early as possible in your batch — the clock starts when credits are granted, not when you start using them

---

## What Credits Cover

### Eligible Services (200+)

| Category | Example Services |
|---|---|
| **Compute** | Amazon EC2, Lambda, ECS, EKS, Fargate |
| **AI/ML** | Amazon Bedrock, SageMaker, Rekognition, Comprehend, Polly, Transcribe, Textract |
| **AI Silicon** | Trainium instances (trn1, trn2), Inferentia instances (inf1, inf2) |
| **GPU Compute** | EC2 P5 (H100), Capacity Blocks for ML (up to 512 H100 GPUs) |
| **Storage** | Amazon S3, EBS, EFS, Glacier |
| **Databases** | Amazon RDS, DynamoDB, Aurora, ElastiCache, Redshift |
| **Networking** | CloudFront, VPC, Route 53 (DNS only, not domains), API Gateway |
| **Analytics** | Athena, EMR, Kinesis, OpenSearch |
| **Developer Tools** | CodeBuild, CodePipeline, CloudWatch |

### Not Covered

- AWS Marketplace purchases (except third-party Bedrock models)
- Route 53 domain registrations or transfers
- AWS Professional Services and Training
- Amazon Mechanical Turk
- Reserved Instance or Savings Plan upfront fees
- Cryptocurrency mining workloads
- AWS Managed Services

---

## Amazon Bedrock — Available AI Models

AWS Activate credits can be used on **all** models available through Amazon Bedrock, including third-party foundation models. As of early 2026, these include:

| Provider | Models |
|---|---|
| **Anthropic** | Claude Opus 4.5, Claude Sonnet 4.5, Claude Haiku 4.5, Claude Opus 4, Claude Sonnet 4 |
| **Meta** | Llama 4 (Scout, Maverick), Llama 3.3, Llama 3.2 |
| **Mistral AI** | Mistral Large 3, Magistral Small, Ministral 3 (3B/8B/14B) |
| **Amazon** | Nova 2 Lite, Nova 2 Pro, Nova Premier, Titan models |
| **Cohere** | Command R+, Command R, Embed |
| **AI21 Labs** | Jamba 2 models |
| **Stability AI** | Stable Diffusion 3.5, SDXL |
| **DeepSeek** | DeepSeek R1 |
| **OpenAI** | GPT-OSS (20B, 120B) |

This is a major advantage — prior to April 2024, AWS credits could **not** be used for third-party Bedrock models. Now they can, making the YC package significantly more valuable for AI startups.

---

## Trainium & Inferentia — AWS Custom AI Silicon

A large portion of the YC package ($300K) is specifically earmarked for AWS’s custom AI chips:

| Chip | Use Case | Key Advantage |
|---|---|---|
| **Trainium (trn1)** | Model training | Up to 50% cost savings vs comparable EC2 GPU instances |
| **Trainium2 (trn2)** | Large-scale training | ~50% cheaper than NVIDIA H100 instances with competitive performance |
| **Trainium3** | Next-gen training (3nm) | 2.52 petaflops FP8, 144 GB HBM3e memory |
| **Inferentia (inf1)** | Inference | Up to 40% better price-performance for inference |
| **Inferentia2 (inf2)** | High-throughput inference | Optimized for LLM and generative AI inference |

Trainium and Inferentia require model compilation via the **AWS Neuron SDK**, which supports PyTorch and TensorFlow. Not all model architectures are supported — check the [Neuron documentation](https://awsdocs-neuron.readthedocs-hosted.com/) for compatibility before committing.

---

## Credit Validity & Expiration

- Credits are valid for **2 years from activation** (activation = when credits are applied to your AWS account)
- Credits **cannot be extended** — the AWS Activate team will not grant extensions under any circumstances
- Unused credits are **forfeited** after expiration
- If you have multiple credit grants, AWS applies the **soonest-expiring credits first**
- Credits apply automatically to charges — you cannot earmark them for specific services
- Monitor your balance in the **AWS Billing Console > Credits** page

---

## Stacking with Other Cloud Programs

YC companies can (and should) apply for credits from multiple cloud providers simultaneously:

| Provider | Program | Amount |
|---|---|---|
| **AWS** | Activate for YC (this program) | Up to $500,000 |
| **Google Cloud** | Google for Startups Cloud Program | Up to $350,000 (AI startups) |
| **Microsoft Azure** | Microsoft for Startups Founders Hub | Up to $150,000 |

Additionally, YC provides credits from other partners during batch onboarding (e.g., Stripe Atlas, Twilio, Segment). These are independent of the AWS package.

---

## Additional Tips

- **Apply immediately** when you join your YC batch — the 2-year clock starts at credit activation, not at the end of the batch
- **Set up AWS Budgets and billing alerts** in the Billing Console — once credits run out, charges go directly to your credit card with no warning
- **Use Cost Explorer** to track burn rate and project when credits will run out
- **Trainium/Inferentia are cheaper but less flexible** than NVIDIA GPUs — not all frameworks and model architectures are supported by the Neuron SDK. Test compatibility early
- **Business Support credits** — the Portfolio tier typically includes up to $10,000 in separate AWS Business Support credits, giving you access to 24/7 technical support
- **You cannot reapply** for the same tier. If you received $500K through YC, you cannot apply again through another Activate Provider for the same or lower amount
- **More than 70% of YC companies run on AWS**, and that number rises to 80% for AI/ML-focused companies
- **Credits are non-transferable** — they are tied to your AWS account and cannot be moved to another account or consolidated billing organization

---

**Sources:**
- [AWS Activate Credits Page](https://aws.amazon.com/startups/credits)
- [AWS Activate Credits for Bedrock Announcement](https://aws.amazon.com/blogs/startups/aws-activate-credits-now-accepted-for-third-party-models-on-amazon-bedrock/)
- [AWS $500K Credits for YC Startups — The Register](https://www.theregister.com/2024/04/02/amazon_ai_credits/)
- [AWS Activate Application Guide](https://aws.amazon.com/startups/learn/applying-for-aws-activate-credits-a-step-by-step-guide)
- [AWS Activate FAQ](https://aws.amazon.com/startups/faq)
- [AWS Activate Terms & Conditions](https://aws.amazon.com/activate/terms/)
- [AWS Credits — Eligible Services](https://aws.amazon.com/awscredits/)
- [Amazon Bedrock Supported Models](https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html)
- [AWS Activate Program Guide 2026 — Cloudvisor](https://cloudvisor.co/aws-activate-program/)
- [$200K AWS Credits Guide — Cloudvisor](https://cloudvisor.co/200k-aws-credits/)

## Submission 10: AWS Activate - Portfolio Package ($100,000)

- Category: Startup Programs
- Source Link: https://aws.amazon.com/startups/credits
- Amount: $100,000
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Provider-backed startups can receive **up to $100,000 in AWS Activate Credits** covering 200+ eligible AWS services, including compute (EC2, ECS, EKS), databases (RDS, DynamoDB), storage (S3), AI/ML (SageMaker, Bedrock), and analytics. To qualify, your startup must be affiliated with an **Activate Provider** (accelerator, VC, or angel investor) such as Y Combinator, Techstars, Sequoia, a16z, or AngelList, and you need their **Organization ID** (a unique, case-sensitive code). Credits expire **1-2 years** from the date of issuance (exact duration varies by provider arrangement). Beyond credits, the package includes up to **$10,000 in AWS Business Support credits** (24/7 technical support, architecture reviews), 80 self-paced training lab credits, and access to $800K+ in third-party exclusive offers from partners like Slack, Notion, and MongoDB.

---

## Eligibility Requirements

| Requirement | Detail |
|---|---|
| **Funding stage** | Pre-Series B (self-funded through Series A) |
| **Activate Provider** | Must be affiliated with an approved accelerator, VC, or incubator |
| **Organization ID** | Must have a valid Org ID from your provider |
| **Funding recency** | Most recent funding round must be within the last 12 months (if applicable) |
| **Company age** | Founded within the last 10 years |
| **Funding cap** | Must not have raised more than $100 million total |
| **Prior credits** | Must not have previously received AWS Activate Credits of equal or greater value |
| **Website** | Must have a fully functioning company website |
| **AWS account** | Must be on a Paid Tier Plan (not free tier) |
| **Excluded** | Government entities are not eligible |

**Who qualifies in practice:** If you went through Y Combinator, Techstars, Antler, On Deck, Plug and Play, or similar accelerators, or if your VC is an approved Activate Provider, you likely qualify. Ask your investor or program manager for the Organization ID -- they should know what this is.

**Who does NOT qualify:** Solo developers without a registered company, side projects, companies past Series B, companies older than 10 years, or startups not affiliated with an Activate Provider. For unaffiliated early-stage startups, see the Founders Package ($1,000) instead.

---

## Registration (Step-by-Step)

1. **Get your Organization ID** -- Contact your accelerator, VC, or incubator and ask for the AWS Activate Organization ID. This is a unique, case-sensitive code. Without it, you cannot apply for the Portfolio Package
2. Go to [aws.amazon.com/startups/credits](https://aws.amazon.com/startups/credits) and click **"Apply Now"**
3. **Create an AWS Builder ID** -- use your personal email address. This ID stays with you across organizations. Verify via the confirmation email before proceeding
4. **Complete your profile** -- use your **business email** (matching your startup's domain) for both the Activate profile and the credit application. Using consistent domain information helps validate your application
5. **Select "Portfolio Package"** when prompted to choose between Founders and Portfolio
6. **Link your AWS account** -- connect your startup's primary AWS account to your Builder ID. If you don't have an AWS account yet, create one using your business email, then return to link it
7. **Provide startup details** -- fill in product information, funding details, and your Organization ID. Be specific about your product and AWS usage plans. Providing inaccurate information will get your application denied
8. **Submit** and wait **7-10 business days** for a decision. Track your status on the Credit Application Status page
9. Once approved, credits appear in your AWS Billing Console within **3-4 hours**

**Important caveats:**
- Use a **business email** (not Gmail/Yahoo). Generic email domains can delay or disqualify your application
- Your **website must be live** with a clear product description and team details. Incomplete or placeholder websites will get you rejected
- You need **administrator permissions** on your AWS account (root user or IAM admin)
- Apply **within 12 months** of your most recent funding date or you lose eligibility
- Time your application carefully -- credits start expiring from the date of issuance, not from when you first use them

---

## What the Credits Cover

### Eligible Services (200+)

| Category | Key Services |
|---|---|
| **Compute** | EC2, Lambda, ECS, EKS, Fargate, Lightsail |
| **Storage** | S3, EBS, EFS, Glacier |
| **Databases** | RDS, DynamoDB, Aurora, ElastiCache, Redshift |
| **AI/ML** | SageMaker, Bedrock, Rekognition, Comprehend, Polly, Transcribe, Translate |
| **Analytics** | Athena, EMR, Kinesis, Glue, QuickSight |
| **Networking** | VPC, CloudFront, Route 53 (hosting only), API Gateway, ELB |
| **DevOps** | CodeBuild, CodeDeploy, CodePipeline, CloudWatch |
| **Security** | GuardDuty, AWS Config, WAF, Secrets Manager |
| **Support** | Developer Support, Business Support |

### Amazon Bedrock (Third-Party AI Models)

Credits explicitly cover third-party foundation models on Amazon Bedrock, giving you access to:
- **Anthropic** (Claude family)
- **Meta** (Llama models)
- **Mistral AI**
- **AI21 Labs** (Jamba models)
- **Cohere** (Command, Embed)
- **Stability AI** (image generation)
- **Amazon** (Titan models)

This is a significant benefit -- Bedrock API calls from third-party models are normally billed through AWS Marketplace, but Activate Credits cover this usage.

### NOT Covered

- AWS Marketplace purchases (except Bedrock third-party models)
- Amazon Mechanical Turk
- AWS Managed Services
- AWS Professional Services
- AWS Training and Certification
- Route 53 domain registration/transfer
- Upfront fees for Savings Plans and Reserved Instances
- Cryptocurrency mining
- Reseller accounts

---

## Additional Benefits Beyond Credits

| Benefit | Detail |
|---|---|
| **Business Support credits** | Up to $10,000 for AWS Business Support (24/7 technical support, architecture reviews, cost optimization) |
| **Training credits** | 80 self-paced lab credits covering ML, security, infrastructure, and application development |
| **Activate Exclusive Offers (AEO)** | Up to $800,000 in third-party discounts (Slack, Notion, MongoDB, Auth0, ClickHouse, Amplitude, Confluent, CrowdStrike, and more) |
| **Startup Showcase** | Directory listing to share your product with investors and peers |
| **Architecture reviews** | Access to AWS specialists for infrastructure design and scaling guidance |
| **Amazon Q Developer** | AI-powered coding assistance |

---

## Credit Mechanics & Gotchas

- **Auto-applied:** Credits apply automatically to all eligible charges each month. You cannot select or prioritize which services consume credits first
- **No retroactive coverage:** Credits only offset new charges going forward, not past bills
- **Expiration is strict:** AWS will **not** extend the expiration date under any circumstances. Unused credits are permanently forfeited
- **Expiry reminders:** AWS sends email alerts at 60 days before, 30 days before, and after expiry, plus at 75% and 100% usage
- **Post-expiry billing:** When credits run out or expire, standard AWS rates kick in immediately for all running services. Set up **budget alerts** in the Billing Console
- **Non-transferable:** Credits cannot be shared with or passed to end customers
- **Reapplication possible:** If you previously received a smaller amount (e.g., $10,000) and now qualify for $100,000, you can reapply and receive the difference ($90,000)
- **Support downgrade:** If your Business Support credits expire before your service credits, you automatically drop to Basic Support

---

## Comparison: Founders vs. Portfolio vs. Generative AI Tier

| Feature | Founders | Portfolio | Generative AI |
|---|---|---|---|
| **Credits** | $1,000 | Up to $100,000 | Up to $300,000 |
| **Provider required** | No | Yes (Org ID) | Yes (selective) |
| **Target stage** | Pre-funding / early | Pre-Series B | AI model builders |
| **Business Support** | Not included | Up to $10,000 | Included |
| **Duration** | 1 year | 1-2 years | 1-2 years |
| **Architecture guidance** | No | Yes | Yes + dedicated mentors |

The **$300,000 Generative AI tier** is restricted to startups building foundation models (not API wrappers). It requires a partnership through the AWS Generative AI Accelerator, a 10-week program with dedicated mentorship.

---

## Additional Tips

- **Timing matters:** Don't apply the day you get accepted to an accelerator. Wait until you're ready to ramp up AWS usage so credits don't expire before you need them. But don't wait more than 12 months from your funding date
- **Founders Package first:** If you're just starting out without a provider, apply for the $1,000 Founders Package immediately. You can upgrade to Portfolio later when you join an accelerator or raise funding
- **Check your provider:** AWS does not publish a full list of Activate Providers. Always ask your investor/accelerator directly if they are an approved provider and request the Org ID
- **AWS Free Tier stacks:** The AWS Always Free Tier (Lambda 1M requests, DynamoDB 25 GB, S3 limited, etc.) continues to work alongside Activate Credits. Free tier usage does not consume credits
- **Set billing alarms:** Configure AWS Budgets and CloudWatch billing alerts immediately after activation. Startups have been surprised by large bills when credits expire with services still running
- **Business email required:** Applications from personal Gmail, Yahoo, or Hotmail addresses are frequently rejected. Use your company domain email
- **Over 350,000 startups** have joined AWS Activate since 2013, with over $8 billion in credits distributed globally

---

**Sources:**
- [AWS Activate Credits Page](https://aws.amazon.com/startups/credits)
- [Applying for AWS Activate Credits: Step-by-Step Guide](https://aws.amazon.com/startups/learn/applying-for-aws-activate-credits-a-step-by-step-guide)
- [Everything You Need to Know About AWS Activate Credits](https://aws.amazon.com/startups/learn/everything-you-need-to-know-about-aws-activate-credits)
- [AWS Activate Terms & Conditions](https://aws.amazon.com/activate/terms/)
- [AWS Activate Providers](https://aws.amazon.com/activate/portfolio-detail/)
- [AWS Startups FAQ](https://aws.amazon.com/startups/faq)
- [AWS Activate Exclusive Offers](https://aws.amazon.com/startups/offers)
- [AWS Activate Credits for Bedrock Models](https://aws.amazon.com/blogs/startups/aws-activate-credits-now-accepted-for-third-party-models-on-amazon-bedrock/)
- [AWS Generative AI Accelerator](https://aws.amazon.com/startups/programs/generative-ai)
- [AWS Activate Program Guide 2026 (Cloudvisor)](https://cloudvisor.co/aws-activate-program/)
- [AWS Startup Credits Guide (StormIT)](https://www.stormit.cloud/blog/aws-startup-credits/)

## Submission 9: AWS Activate - Generative AI Tier ($300,000)

- Category: Startup Programs
- Source Link: https://aws.amazon.com/startups/credits
- Amount: $300,000
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

AWS Activate offers an elevated credit tier of **up to $300,000** for AI startups building foundation models. These credits are redeemable across 200+ AWS services including **Amazon Bedrock** (with third-party models from Anthropic, Meta, Mistral, Cohere, and others), plus compute-heavy workloads on **AWS Trainium** (training) and **AWS Inferentia** (inference) chips. This is not the standard Activate Portfolio ($100K max) — it is a separate, higher allocation within the Activate framework, restricted to startups genuinely building foundational AI technology. Nomination from a recognized VC partner is typically required.

---

## How This Differs from Standard AWS Activate

AWS Activate has three tiers. The Generative AI tier sits above the standard Portfolio package:

| Tier | Credits | Who It's For |
|---|---|---|
| **Founders** | Up to $1,000 | Early-stage, self-funded startups |
| **Portfolio** | Up to $100,000 | VC-backed or accelerator-affiliated startups |
| **Generative AI** | Up to $300,000 | Startups building foundation models (not API wrappers) |

The Generative AI tier is not a separate application form — it is an elevated allocation within the Portfolio framework. Your VC or accelerator partner, combined with your demonstrated need for compute-heavy AI workloads, determines whether you qualify for the higher amount. Credits do not stack: if you previously received $100K in Portfolio credits, qualifying for the GenAI tier would grant you an additional $200K (the difference).

---

## Eligibility

**Hard requirements:**
- Pre-Series B startup (most recent funding round within the last 12 months, if applicable)
- Founded within the last 10 years
- Active AWS account on a paid-tier plan
- Fully functioning company website with a clear product description
- Corporate email address (personal Gmail/Yahoo will get you rejected)
- Valid Organization ID from a recognized VC, accelerator, or AWS Activate Provider

**GenAI-tier specific requirements:**
- You must be **building foundation models** — fine-tuning open-source models, pre-training your own FMs, or providing critical infrastructure in the generative AI stack
- Mere API wrappers around GPT-4, Claude, or other hosted models do **not** qualify
- You must demonstrate a need for **compute-heavy workloads** (e.g., clusters of P5 or G5 instances, Trainium clusters)
- Typically requires **nomination from a high-profile VC partner** — top-tier firms like Y Combinator, Sequoia, and a16z can secure higher credit allocations
- U.S. startups need an EIN; non-U.S. startups may use local tax IDs
- Government entities are excluded

---

## Application Process (Step-by-Step)

1. **Secure an Organization ID** — contact your VC, accelerator, or startup program to obtain a valid Org ID. This is a unique, case-sensitive code. Not all providers offer the same tier — ask your provider what credit amount they can support before applying
2. **Set up your AWS account** — sign up at [aws.amazon.com](https://aws.amazon.com) with a corporate email matching your website domain. Confirm email, credit card, and phone number
3. **Create an AWS Builder ID** — this links your application to your AWS account. Ensure the email is not tied to another Builder ID (AWS rejects duplicates)
4. **Go to the AWS Activate application page** at [aws.amazon.com/startups/credits](https://aws.amazon.com/startups/credits)
5. **Select Activate Portfolio** and fill out the application with: company name, website, description, AWS account ID, Org ID, and funding details
6. **Describe your AI workloads** — be specific about your foundation model work, compute needs (instance types, training pipeline), and how you plan to use Bedrock, Trainium, or Inferentia. This description directly affects your credit allocation
7. **Submit and wait** — AWS reviews applications within **7-10 business days**, but the full approval process for higher-tier credits can take **6-10 weeks**
8. **Credits appear in your AWS Billing Dashboard** upon approval

**Common rejection pitfalls:**
- Using a personal email address (Gmail, Yahoo)
- Website that does not clearly describe your product
- Inaccurate or vague funding/product information
- Previously received Activate credits of equal or greater value
- Applying without a valid Org ID

---

## What Credits Cover

Credits are redeemable on **200+ eligible AWS services**, including:

| Category | Services |
|---|---|
| **AI/ML** | Amazon Bedrock (including third-party models), SageMaker, Trainium instances, Inferentia instances |
| **Compute** | EC2 (including P5, G5 GPU instances), Lambda, Fargate, ECS, EKS |
| **Storage** | S3, EBS, EFS |
| **Databases** | RDS, DynamoDB, Aurora, ElastiCache |
| **Analytics** | Redshift, Kinesis, Glue |
| **Networking** | CloudFront, VPC |

### Amazon Bedrock Models (covered by credits)

Activate credits can offset costs for third-party foundation models on Bedrock, including:

| Provider | Models |
|---|---|
| **Anthropic** | Claude Opus 4.5, Claude Sonnet 4.5, Claude Sonnet 4, Claude Haiku 4.5 |
| **Meta** | Llama 4, Llama 3.3, Llama 3.2, Llama 3.1 |
| **Mistral AI** | Mistral Large and others |
| **AI21 Labs** | Jamba 1.5 Mini, Jamba 1.5 Large |
| **Cohere** | Command, Embed, Rerank |
| **Amazon** | Amazon Nova, Amazon Titan |
| **Others** | DeepSeek, OpenAI, Qwen, Stability AI, Writer |

Anthropic models require a one-time use case submission (access granted immediately after submission). Most other models are available with no approval needed.

---

## What Credits Do NOT Cover

- **AWS Marketplace** purchases (except Bedrock third-party models)
- **Route 53** domain registrations
- **Reserved Instance** upfront fees ("All Upfront" RI payments)
- **AWS Managed Services**
- Prior/retroactive bills — credits only apply to new usage
- Credits are applied automatically across all eligible services; you cannot manually allocate them to specific services

---

## Trainium & Inferentia: Purpose-Built AI Chips

The GenAI tier is specifically designed for startups leveraging AWS's custom silicon:

| Chip | Purpose | Advantage |
|---|---|---|
| **AWS Trainium** (Trn1 instances) | Model training | Up to 50% cost savings vs comparable EC2 GPU instances |
| **AWS Inferentia** (Inf2 instances) | Model inference | Up to 40% better price-performance vs GPU alternatives |

Startups can combine approaches: train models on Trainium, deploy inference with Inferentia, and use Bedrock for rapid prototyping with hosted foundation models.

---

## Credit Expiration & Lifecycle

- Credits expire **2 years from activation** (use-it-or-lose-it)
- AWS **cannot** extend expiration dates, even if credits remain unused
- AWS sends email reminders at **60 days before**, **30 days before**, and **after expiry**
- Usage alerts are sent when credits reach **75%** and **100%** consumption
- After credits expire, you are charged standard pay-as-you-go rates for all running services
- Credits are **non-transferable** — they cannot be passed to end customers

---

## Related AWS Programs for AI Startups

| Program | Credits | Details |
|---|---|---|
| **Generative AI Accelerator** | Up to $1,000,000 | 8-week cohort program, <2% acceptance rate, 40 startups per cohort, includes mentorship and investor access |
| **AWS & Anthropic GenAI Scale Program** | $5-10K immediately + up to $300K after | Requires deployments via Bedrock; must separately qualify for Activate |
| **Activate Founders** | Up to $1,000 | No VC needed, self-funded startups |
| **Activate Portfolio** | Up to $100,000 | VC/accelerator-affiliated startups |

You can stack credits from different cloud providers (AWS + GCP + Azure) simultaneously — there is no exclusivity requirement between providers.

---

## Additional Tips

- **Your VC matters** — top-tier VCs (YC, Sequoia, a16z) routinely secure the full $100K-$300K. Lesser-known accelerators may only offer $25K-$50K. Ask your provider about their tier before applying
- **Application pitfalls** — applying without proper guidance can land you in a lower tier. Once you trigger a lower allocation (e.g., $25K when you qualified for $100K), upgrading is administratively difficult
- **AWS partners can help** — companies like Cloudvisor, nOps, and other AWS consulting partners can guide your application to maximize credit allocation
- **Set up billing alerts immediately** — AWS has no spending caps. If credits run out, you pay full price with no warning beyond the email notifications
- **Batch inference pricing** — Bedrock offers 50% lower pricing for batch inference vs on-demand. Use this to stretch your credits further when running large-scale model evaluations
- **Region availability** — not all Bedrock models or Trainium instances are available in every AWS region. Check [regional availability](https://docs.aws.amazon.com/bedrock/latest/userguide/models-regions.html) before planning your infrastructure
- **Since 2013**, AWS Activate has provided over $7 billion in promotional credits to startups globally

---

**Sources:**
- [AWS Activate Credits](https://aws.amazon.com/startups/credits)
- [AWS Activate Application Guide](https://aws.amazon.com/startups/learn/applying-for-aws-activate-credits-a-step-by-step-guide)
- [Everything About AWS Activate Credits](https://aws.amazon.com/startups/learn/everything-you-need-to-know-about-aws-activate-credits)
- [AWS Activate Credits for Bedrock](https://aws.amazon.com/blogs/startups/aws-activate-credits-now-accepted-for-third-party-models-on-amazon-bedrock/)
- [AWS Generative AI for Startups](https://aws.amazon.com/startups/generative-ai/)
- [AWS Generative AI Accelerator](https://aws.amazon.com/startups/generative-ai/accelerator/)
- [AWS Activate Program Guide 2026 (Cloudvisor)](https://cloudvisor.co/aws-activate-program/)
- [AWS Startup Credits Guide (Cloudvisor)](https://cloudvisor.co/aws-credits-for-startups/)
- [AWS Activate Terms & Conditions](https://aws.amazon.com/activate/terms/)
- [AWS Activate FAQ](https://aws.amazon.com/startups/faq)
- [Amazon Bedrock Supported Models](https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html)
- [Amazon Bedrock Regional Availability](https://docs.aws.amazon.com/bedrock/latest/userguide/models-regions.html)

## Submission 8: AWS Activate - Founders Package ($1,000)

- Category: Startup Programs
- Source Link: https://aws.amazon.com/activate/activate-landing
- Amount: $1,000 in AWS credits + $350 in Developer Support credits
- Difficulty: 3/5 (Medium)
- Expires At: N/A

### Description

Self-funded and bootstrapped startups can get **$1,000 in AWS credits** plus **$350 in Developer Support credits** through the AWS Activate Founders Package. Credits cover 200+ AWS services including compute (EC2, Lambda), storage (S3), databases (RDS, DynamoDB), and AI/ML services like **Amazon Bedrock** and **Amazon SageMaker**. No VC funding required. You need a live company website, a corporate email address, and an AWS account for a company founded in the last 10 years. Credits expire **1 year after activation** and are use-it-or-lose-it.

---

## Registration (Step-by-Step)

1. **Create an AWS account** at [aws.amazon.com](https://aws.amazon.com) if you don't have one already. A credit card is required for the account (standard AWS requirement)
2. **Create an AWS Builder ID** at [profile.aws](https://profile.aws) using your personal email. Verify the email confirmation
3. Go to the [AWS Activate Console](https://console.aws.amazon.com/activate) or [aws.amazon.com/activate](https://aws.amazon.com/activate)
4. **Select the Founders Package** (if you don't have an Org ID from an accelerator/VC, this is your option)
5. **Link your AWS account** to your Builder ID by clicking "Verify your account" in the console. You should see "Accounts linked successfully"
6. **Fill in your startup details** — company name, product description, target market, funding stage (select pre-seed/seed/bootstrapped), and team info
7. **Use your corporate email** (e.g., founder@yourstartup.com) — this must match your company domain. Personal Gmail/Yahoo/Hotmail addresses will almost certainly get you rejected
8. **Review and submit** — double-check all details, especially funding status and email domain consistency
9. Wait **7-10 business days** for a decision. Check your email (including spam) and the Credit Application Status page for updates

**Important caveats:**
- One Founders Package per startup — if you've previously received AWS Activate credits, you're ineligible
- Your website must be live, professional, and clearly describe a software product. "Coming Soon" pages or placeholder sites get rejected
- Your LinkedIn company page should exist and match the information on your application
- Don't position yourself as a consulting/agency company — AWS Activate is for product companies only

---

## What's Included

| Benefit | Details |
|---|---|
| **AWS Credits** | $1,000 toward 200+ eligible AWS services |
| **Developer Support Credits** | $350 toward AWS Developer Support (unlimited email-based technical support cases) |
| **Training Resources** | Self-paced technical guides, startup best practices, MVP building guides |
| **AWS Activate Offers** | Exclusive discounts from partners like Datadog, Zendesk, Stripe, and others |
| **AWS Startup Lofts** | Access to physical co-working spaces in select cities (where available) |

---

## AI & ML Services Covered by Credits

AWS Activate credits can be used for all major AI/ML services on AWS:

| Service | What It Does |
|---|---|
| **Amazon Bedrock** | Managed access to foundation models (Claude, Llama, Mistral, Amazon Nova, and more) via a single API |
| **Amazon SageMaker** | Build, train, and deploy ML models with managed infrastructure |
| **Amazon Rekognition** | Image and video analysis (object detection, facial analysis, text extraction) |
| **Amazon Comprehend** | NLP — sentiment analysis, entity recognition, key phrase extraction |
| **Amazon Transcribe** | Speech-to-text |
| **Amazon Polly** | Text-to-speech |
| **Amazon Translate** | Real-time language translation |
| **Amazon Textract** | Extract text and structured data from documents |
| **Amazon Q** | AI-powered assistant for developers and businesses |

**Bedrock models available** (most require no approval, Anthropic requires a one-time use case submission):
- **Anthropic** — Claude Sonnet 4, Claude Opus 4.5, Claude 3.5 Sonnet, Claude 3 Haiku
- **Meta** — Llama 4, Llama 3.3, Llama 3.2
- **Mistral AI** — Mistral Large 3, Ministral 3
- **Amazon** — Nova Pro, Nova Lite, Nova Micro
- **AI21 Labs** — Jamba models
- **Cohere** — Command R/R+, Embed
- Plus models from OpenAI, Google, Qwen, and others (nearly 100 serverless models total)

---

## What Credits Do NOT Cover

The following services and charges are **excluded** from AWS Activate credits:

- **Amazon Mechanical Turk**
- **AWS Managed Services**
- **AWS Marketplace** (except third-party foundation models on Amazon Bedrock)
- **AWS Professional Services**
- **AWS Training and Certification** (paid courses)
- **Route 53 domain registration or transfer**
- **Cryptocurrency mining**
- **Upfront fees for Savings Plans and Reserved Instances**
- **Reseller accounts**

---

## Common Rejection Reasons

AWS Activate applications are frequently rejected. Here are the top reasons and how to avoid them:

1. **Personal email address** — Using Gmail, Yahoo, or Hotmail instead of a corporate domain email. This is the number one cause of rejection
2. **No live website** — Your company website must be fully functional and clearly describe your product. Placeholder or "coming soon" pages don't count
3. **Looking like a services company** — If your website says "consulting", "agency", or "we build software for clients", you'll be rejected. AWS Activate is for product startups
4. **Incomplete or inconsistent information** — Your application details must match your website, LinkedIn, and any other public information
5. **Previously received credits** — If you or your company have already received AWS Activate credits under any package
6. **No LinkedIn company page** — AWS checks for a LinkedIn presence that matches your application

If rejected, you can re-apply after fixing the issues. Many startups succeed on their second or third attempt.

---

## Credit Expiration & Billing

- Credits expire **1 year** after they are issued — no extensions
- Credits apply **automatically** each month to eligible charges. You cannot choose which service they apply to
- **After credits are exhausted or expire**, your credit card on file will be charged for any ongoing usage at standard AWS rates
- AWS sends email alerts at **75% exhaustion**, **100% exhaustion**, and at **60 days**, **30 days**, and **0 days** before expiration
- Set up **AWS Budget Alerts** in the Billing Console to avoid surprise charges
- Credits do **not** apply retroactively to past bills

---

## Upgrading to Portfolio Package

If your startup later joins an accelerator, incubator, or raises VC funding, you can apply for the **Activate Portfolio Package** which offers up to **$100,000 in credits** (valid for 2 years) and up to **$10,000 in Business Support credits**. You'll need an Org ID from your Activate Provider (accelerator/VC). Portfolio credits are added as a "top-up" — they don't replace your Founders credits.

For AI-focused startups building foundation models, AWS offers a special tier of up to **$300,000 in credits**, though this typically requires nomination from a recognized VC partner.

---

## Additional Tips

- **No credit card hold** — unlike some cloud providers, AWS doesn't place a temporary authorization hold during signup. But your card will be charged if you exceed credits
- **AWS Free Tier still applies** — even without Activate credits, AWS has an always-free tier (750 hours EC2 t2.micro/month for 12 months, 25 GB DynamoDB, 1M Lambda requests/month, etc.). These free tier allocations are separate from your Activate credits
- **Monitor your spending** — $1,000 goes fast if you're running GPU instances or heavy Bedrock API usage. A single ml.p3.2xlarge SageMaker instance costs ~$3.83/hour
- **Credits are non-transferable** — you cannot sell, trade, or move credits to another AWS account
- **Consolidated billing** — if you use AWS Organizations, credits on the payer account can apply to linked accounts, but be careful about unintended usage
- **Apply early** — the 1-year clock starts when credits are issued, not when you first use them

---

**Sources:**
- [AWS Activate Credits](https://aws.amazon.com/startups/credits)
- [Applying for AWS Activate Credits: Step-by-Step Guide](https://aws.amazon.com/startups/learn/applying-for-aws-activate-credits-a-step-by-step-guide)
- [Everything You Need to Know About AWS Activate Credits](https://aws.amazon.com/startups/learn/everything-you-need-to-know-about-aws-activate-credits)
- [AWS Activate FAQ](https://aws.amazon.com/startups/faq)
- [AWS Activate Credits Now Accepted for Bedrock Models](https://aws.amazon.com/blogs/startups/aws-activate-credits-now-accepted-for-third-party-models-on-amazon-bedrock/)
- [AWS Activate Program 2026 Guide (Cloudvisor)](https://cloudvisor.co/aws-activate-program/)
- [AWS Startup Credits Eligibility Guide (SquareOps)](https://squareops.com/knowledge/who-is-eligible-for-aws-startup-credits-and-how-to-apply-2025-guide/)
- [Amazon Bedrock Pricing](https://aws.amazon.com/bedrock/pricing/)

## Submission 7: Anthropic Startup Program (Anthology Fund) - $25,000

- Category: Startup Programs
- Source Link: https://claude.com/programs/startups
- Amount: $25,000 in Claude API credits
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Early-stage AI startups can receive **$25,000 in Claude API credits** through Anthropic's startup ecosystem, which consists of two distinct programs: the **Anthology Fund** (a $100M venture fund with Menlo Ventures that also invests $100K+ into companies) and the broader **VC Partner Program** (API credits distributed through partner VC firms). Both programs include priority rate limits, technical support from Anthropic, and access to exclusive founder events. Credits are applied to Anthropic's API and cover all Claude models. No credit card is required for the credits themselves — Anthropic uses a prepaid credit model, so you will not be auto-charged when credits run out.

---

## Two Programs, Often Confused

Anthropic runs two separate startup-facing programs that are frequently mixed up:

| | Anthology Fund | VC Partner Program |
|---|---|---|
| **What it is** | $100M venture fund (Menlo Ventures + Anthropic) | Credits program through partner VC firms |
| **Investment** | $100K+ equity investment (pre-seed to Series A) | No investment — credits only |
| **Credits** | $25,000 in Claude API credits | Varies (typically $1,000–$25,000) |
| **How to apply** | Directly at menlovc.com/anthology-fund-application | Through your VC's unique link |
| **Selection** | Highly competitive (18 selected from thousands) | Broader access if your VC is a partner |

The entry on this page primarily covers the **Anthology Fund**, but the VC Partner Program is the more accessible path for most startups.

---

## Anthology Fund — Registration (Step-by-Step)

1. Go to [menlovc.com/anthology-fund-application](https://menlovc.com/anthology-fund-application/)
2. Fill out the embedded Airtable application form with details about your startup, team, AI use case, and traction
3. Submit — there is **no deadline**, applications are reviewed on a rolling basis
4. Menlo and Anthropic use AI-assisted scoring to rank applications, then manually review top candidates
5. Expect to hear back within **~2 weeks**
6. If shortlisted, you will be invited to **1–2 Zoom calls** with the Menlo/Anthropic team
7. If selected, sign compliance affidavits and publicity releases within **14 calendar days** or forfeit benefits
8. Receive your credits, onboarding, and venture support

**Important:**
- This is a venture investment — Menlo writes checks of $100K+ into your company
- You do **not** need to be building exclusively on Claude, but it helps
- The first cohort selected only 18 companies from thousands of applicants
- Focus areas: AI infrastructure, healthcare, legal, education, scientific research, consumer AI, trust/safety tooling

---

## VC Partner Program — How to Get Credits Without the Fund

If you are not seeking venture investment but want API credits:

1. Check with your existing VC investors whether they are an **Anthropic partner firm**
2. If yes, ask them for their **unique program application link**
3. Apply through that link — your VC submits on your behalf or gives you a referral code
4. Anthropic reviews and grants credits to qualifying portfolio companies

**Known/likely partner firms** (Anthropic does not publish a full list):
- Menlo Ventures
- Lightspeed Ventures
- Accel
- a16z (Andreessen Horowitz)
- Y Combinator
- Spark Capital
- Sequoia Capital

VCs can apply to become partners at [claude.com/contact-sales/vc-partner](https://claude.com/contact-sales/vc-partner).

---

## What You Get (Both Programs)

- **$25,000 in Claude API credits** (Anthology Fund; VC Partner Program amounts vary)
- **Priority rate limits** — the highest tier of publicly available rate limits, critical for production apps
- **Technical support** — direct access to Anthropic's engineering team, office hours, and advance notice of API changes
- **Exclusive events** — quarterly deep dives and biannual demo days with Anthropic leadership (CPO Mike Krieger, President Daniela Amodei)
- **Community** — networking with other AI founders in the program
- **Anthology Fund only:** equity investment ($100K+), fractional workspace in Menlo's SF and Menlo Park offices, credits from Menlo's infrastructure partners

---

## Available Claude Models

Credits can be used across all Claude models available via the Anthropic API:

| Model | Input (per 1M tokens) | Output (per 1M tokens) | Best for |
|---|---|---|---|
| **Claude Opus 4.5** | $5.00 | $25.00 | Most capable, complex reasoning |
| **Claude Sonnet 4.5** | $3.00 | $15.00 | Balanced cost/performance |
| **Claude Haiku 4.5** | $1.00 | $5.00 | Fast, high-volume, cost-efficient |
| Claude Opus 4 (legacy) | $15.00 | $75.00 | Previous generation |
| Claude Sonnet 4 (legacy) | $3.00 | $15.00 | Previous generation |
| Claude Haiku 3.5 (legacy) | $0.80 | $4.00 | Budget option |

**How far $25,000 goes:** Using Claude Sonnet 4.5, $25K buys roughly **8.3 billion input tokens** or **1.67 billion output tokens**. With Haiku 4.5, you get **25 billion input tokens** — enough for serious prototyping and early production.

**Cost-saving features:** Batch API gives a 50% discount on async requests. Prompt caching reduces repeat input costs by up to 90%. Extended thinking tokens are billed at the standard output rate.

---

## Credit Expiration & Billing

- Credits **expire 12 months** from the grant date (typical; check your specific award terms)
- Anthropic uses a **prepaid credit model** — when credits run out, API access stops. There is **no automatic charge** to a credit card
- Auto-reload is available but is **opt-in only** — you must explicitly enable it
- All credits are **non-refundable** and expiration dates cannot be extended
- If you are not yet building with Claude, **do not apply too early** — the 12-month clock starts at grant, not at first usage

---

## Eligibility & Restrictions

**Who qualifies:**
- Early-stage startup (pre-seed through Series A for Anthology Fund)
- Building AI-powered products or solutions
- Backed by a partner VC firm (for VC Partner Program) or applying directly (for Anthology Fund)

**Who does not qualify:**
- Companies established or operating in Belarus, China, Cuba, Iran, Myanmar, North Korea, Russia, Sudan, Syria, Crimea, Donetsk, or Luhansk
- Anthropic employees, officers, directors, or their immediate families
- Companies that cannot comply with U.S. export control laws

**Compliance requirements:**
- Selected awardees must sign compliance affidavits and publicity releases within 14 calendar days
- Tax responsibilities fall on the recipient — API credits may have tax implications
- Anthropic can cancel or modify the program at any time
- Liability is capped at the lesser of the actual credit value or $1,000

---

## Alternative Ways to Access Claude Credits

If you do not qualify for the Anthology Fund or VC Partner Program:

- **AWS Activate** — startups on AWS can access Claude via Amazon Bedrock using their AWS credits (up to $100K for qualifying startups)
- **Google for Startups Cloud Program** — Scale AI tier includes $10,000 in credits usable toward Anthropic models in Vertex AI Model Garden
- **Microsoft for Startups Founders Hub** — Azure credits can be used for some AI services, but Anthropic Claude models on Azure are billed separately as third-party models (not covered by Azure startup credits)
- **Anthropic API free tier** — new API accounts get a small amount of free credits ($5) to test with

---

## Additional Tips

- **Timing matters** — credits expire 12 months from grant. Apply when you are ready to build, not when you are still exploring ideas
- **Rate limits matter more than credits** — priority rate limits determine whether you can scale a production app. This is often more valuable than the dollar amount
- **You do not need to use Claude exclusively** — Anthology Fund companies are not required to build on Claude, though it is preferred
- **The Anthology Fund is an investment, not a grant** — Menlo Ventures takes equity. If you only want credits without dilution, the VC Partner Program is the better path
- **Response time is fast** — expect ~2 weeks for Anthology Fund decisions, compared to months at typical VCs
- **Region note** — while the program is technically global, Menlo Ventures is US-focused and the co-working spaces are in San Francisco and Menlo Park. Non-US startups can apply but may get less hands-on support

---

**Sources:**
- [Anthropic Startup Program](https://claude.com/programs/startups)
- [Anthropic Startup Program Official Terms](https://www.anthropic.com/startup-program-official-terms)
- [Anthropic VC Partner Program Official Terms](https://www.anthropic.com/vc-partner-program-official-terms)
- [Anthology Fund (Menlo Ventures)](https://menlovc.com/anthology-fund/)
- [Anthology Fund Application](https://menlovc.com/anthology-fund-application/)
- [Anthropic + Menlo Ventures Launch Announcement](https://www.anthropic.com/news/anthropic-partners-with-menlo-ventures-to-launch-anthology-fund)
- [Anthology Fund First Cohort (TechCrunch)](https://techcrunch.com/2024/12/18/menlo-ventures-and-anthropic-have-picked-the-first-18-startups-for-their-100m-fund/)
- [Claude API Pricing](https://platform.claude.com/docs/en/about-claude/pricing)
- [Anthropic API Billing Help](https://support.anthropic.com/en/articles/8977456-how-do-i-pay-for-my-api-usage)
- [VC Partner Application](https://claude.com/contact-sales/vc-partner)

## Submission 6: Anthropic Fellows Program 2026 - Funded AI Safety Research

- Category: Education & Student
- Source Link: https://alignment.anthropic.com/2025/anthropic-fellows-program-2026
- Amount: $3,850/week stipend + ~$15,000/month compute budget
- Difficulty: 5/5 (Very Hard)
- Expires At: N/A

### Description

Four-month funded research fellowship for engineers and researchers working on Anthropic's highest-priority AI safety and security questions. Fellows receive a **$3,850/week stipend** (or equivalent in GBP/CAD), approximately **$15,000/month in compute funding**, direct 1:1 mentorship from Anthropic researchers, and access to shared workspaces in Berkeley or London. No PhD, prior ML experience, or published papers required. Over 80% of previous fellows produced publishable research, and over 40% joined Anthropic full-time afterward.

---

## Correction Notice

The original seed entry listed the stipend as $2,100/week and compute budget as $10,000/month. The actual figures from Anthropic's official announcement are **$3,850 USD/week** and **~$15,000/month** in compute funding.

---

## Registration (Step-by-Step)

1. Go to the [Anthropic Fellows Program announcement](https://alignment.anthropic.com/2025/anthropic-fellows-program-2026/) and read the program details
2. Decide which track fits you best: **AI Safety Fellow** or **AI Security Fellow** (apply to only one)
3. Click **"Apply here"** on the announcement page — this redirects to [Constellation's application portal](https://job-boards.greenhouse.io/anthropic/jobs/5023394008) (Constellation is Anthropic's official recruiting partner)
4. Fill out the application form with your background, research interests, and motivation
5. Applications go through **initial screening**, followed by **technical assessments and interviews**, then a **research discussion round**
6. If accepted, Anthropic mentors pitch project ideas to you, and you choose and shape your project in collaboration with your mentor
7. The fellowship begins at the cohort start date (May or July 2026)

**Important:**
- Applications for the **May 2026 cohort are closed**
- Applications are accepted on a **rolling basis** for July 2026 and beyond
- The initial application deadline was **January 12, 2026** for the first round, but rolling applications continue
- **No visa sponsorship** — you must already have (or independently obtain) full-time work authorization in the US, UK, or Canada
- You can participate **remotely** from anywhere within the US, UK, or Canada

---

## Two Fellowship Tracks

| Track | Focus | Application Link |
|---|---|---|
| **AI Safety Fellow** | Scalable oversight, adversarial robustness, model organisms, mechanistic interpretability, model welfare | [Apply here](https://job-boards.greenhouse.io/anthropic/jobs/5023394008) |
| **AI Security Fellow** | Defensive AI for cybersecurity, vulnerability research, securing code and infrastructure with AI | [Apply here](https://job-boards.greenhouse.io/anthropic/jobs/5030244008) |

Both tracks share the same compensation, duration, and structure. Apply to only the one that best matches your interests and background.

---

## Research Areas

### AI Safety Track
- **Scalable oversight** — methods to supervise AI systems as they become more capable
- **Adversarial robustness and AI control** — preventing misuse and maintaining alignment under adversarial conditions
- **Model organisms** — studying misalignment in controlled settings
- **Mechanistic interpretability** — understanding internal model representations and circuits
- **Model welfare** — investigating the moral status of AI systems

### AI Security Track
- **Defensive AI** — using AI models to discover vulnerabilities and secure code
- **Blockchain smart contract exploits** — AI-assisted security research
- **Control evaluations** — testing how well safety measures hold up
- **Offensive security research** — pentesting and vulnerability research applied to AI systems

---

## What You Get

| Benefit | Details |
|---|---|
| **Weekly stipend** | $3,850 USD / £2,310 GBP / $4,300 CAD |
| **Compute funding** | ~$15,000/month for research expenses |
| **Mentorship** | 1:1 guidance from senior Anthropic researchers |
| **Workspace** | Shared offices in **Berkeley, CA** or **London, UK** |
| **Remote option** | Available for US, UK, or Canada residents |
| **Duration** | 4 months full-time |
| **Research output** | Aim to produce a publishable paper |
| **Career pipeline** | Over 40% of fellows received full-time Anthropic offers |

---

## Eligibility & Candidate Profile

**Required:**
- Fluent **Python programming** ability
- Available to work **full-time for 4 months**
- **Work authorization** in the US, UK, or Canada (no visa sponsorship)
- Strong technical background in computer science, mathematics, physics, cybersecurity, or related fields
- Genuine interest in reducing catastrophic risks from advanced AI

**Valued but not mandatory:**
- Prior ML research experience
- Published papers or PhD
- Experience with deep learning frameworks
- Open-source contributions (especially in LLM- or security-adjacent repos)
- For Security track: pentesting, CVE reports, bug bounty experience

**Key message from Anthropic:** They care much more about your ability to execute on research than about credentials. Successful fellows have come from physics, mathematics, computer science, cybersecurity, and other quantitative backgrounds.

---

## Mentors

### Safety Track
Mentors are senior Anthropic researchers working on alignment and interpretability (specific names listed on the announcement page).

### Security Track
Potential mentors include **Nicholas Carlini**, **Keri Warr**, **Evyatar Ben Asher**, **Keane Lucas**, and **Newton Cheng**.

Mentors pitch project ideas to fellows, who then choose and shape their project collaboratively.

---

## Cohort Timeline

| Cohort | Start Date | Application Status |
|---|---|---|
| May 2026 | May 2026 | **Closed** |
| July 2026 | July 2026 | **Open** (rolling basis) |
| Future cohorts | TBD | **Open** (rolling basis) |

---

## Track Record

- Over **80% of fellows** in the first cohort produced published papers
- Over **40% joined Anthropic full-time** after the fellowship
- Many others were supported to work full-time on AI safety at other organizations
- Published work includes research on agentic misalignment, subliminal learning, rapid response to ASL3 jailbreaks, and open-source circuits
- One fellow developed a method for rapid response to new ASL3 jailbreaks that became a key component of Anthropic's deployment safeguards

---

## Additional Tips

- **This is not a typical internship** — it is a structured research program with the expectation of producing publishable work. Come with genuine research motivation, not just a line on your resume
- **Constellation manages logistics** — they handle applications, interviews, and run the Berkeley workspace. Expect communication from Constellation, not Anthropic, during the application process
- **No credit card or payment required** — this is a fully funded fellowship; you receive money, not spend it
- **Compare with similar programs** — the [Astra Fellowship by Constellation](https://www.constellation.org/programs/astra-fellowship) is a related program worth exploring if this one doesn't fit
- **Rolling applications** — don't wait until the last minute, but if you missed the January deadline, you can still apply for July 2026 and future cohorts
- **Remote is fine** — you don't have to relocate to Berkeley or London, but workspace access is available if you want it

---

**Sources:**
- [Anthropic Fellows Program 2026 Announcement](https://alignment.anthropic.com/2025/anthropic-fellows-program-2026/)
- [AI Safety Fellow Application (Greenhouse)](https://job-boards.greenhouse.io/anthropic/jobs/5023394008)
- [AI Security Fellow Application (Greenhouse)](https://job-boards.greenhouse.io/anthropic/jobs/5030244008)
- [Introducing the Anthropic Fellows Program (2024)](https://alignment.anthropic.com/2024/anthropic-fellows-program/)
- [Anthropic Fellows Program Coverage (AImpactful)](https://aimpactful.com/anthropic-opens-applications-for-ai-safety-fellows-program-2026/)
- [Anthropic Announcement on X](https://x.com/AnthropicAI/status/1999233249579794618)

## Submission 5: Anthropic Claude API - $5 Free Credits

- Category: AI API Free Tiers
- Source Link: https://console.anthropic.com/
- Amount: $5 free credits (no credit card required)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

New developer accounts on [console.anthropic.com](https://console.anthropic.com) can claim **$5 in free API credits** by verifying a phone number during signup. No credit card is required. The credits work across all available Claude models, including Haiku, Sonnet, and Opus. With $5 you can run roughly **5 million input tokens on Haiku 3** or **1 million input tokens on Opus 4.5** — enough for meaningful prototyping. The credits **expire 14 days after activation**, so claim them only when you are ready to build.

---

## Registration (Step-by-Step)

1. Go to [console.anthropic.com](https://console.anthropic.com)
2. Click **"Sign Up"** — you can register with Google, Microsoft, Apple, or a plain email and password
3. **Verify your email** — check your inbox and click the confirmation link
4. **Phone verification** — enter a mobile phone number capable of receiving SMS. You will receive a six-digit code (may take up to 5 minutes)
5. After successful verification, a banner appears at the top of the Console dashboard
6. Click the **"Claim"** button on the banner to activate your $5 credit
7. Go to **Settings > API Keys** and generate your first API key

**Important:**
- VoIP numbers, Google Voice, app-generated numbers, and landlines are **not accepted**
- US-based phone numbers work reliably; UK numbers are explicitly excluded; other regions have variable eligibility
- You **cannot change your phone number** after verification — use a long-term number
- One free credit claim per account; creating multiple accounts violates the Terms of Service and risks a permanent ban
- The 14-day credit expiration clock starts when you click "Claim", not when you create the account

---

## Available Models

All Claude models are accessible with free credits. Choosing the right model determines how far your $5 goes.

| Model | Input Price | Output Price | $5 Gets You (Input) |
|---|---|---|---|
| **Claude Haiku 3** | $0.25 / MTok | $1.25 / MTok | ~20M tokens |
| **Claude Haiku 3.5** | $0.80 / MTok | $4.00 / MTok | ~6.25M tokens |
| **Claude Haiku 4.5** | $1.00 / MTok | $5.00 / MTok | ~5M tokens |
| **Claude Sonnet 4.5** | $3.00 / MTok | $15.00 / MTok | ~1.67M tokens |
| **Claude Sonnet 4** | $3.00 / MTok | $15.00 / MTok | ~1.67M tokens |
| **Claude Opus 4.5** | $5.00 / MTok | $25.00 / MTok | ~1M tokens |
| **Claude Opus 4.1** | $15.00 / MTok | $75.00 / MTok | ~333K tokens |

**Tip:** Start prototyping with **Haiku 3 or 3.5** to stretch your credits. Switch to Sonnet or Opus only for tasks that need higher reasoning quality. Opus 4.1 burns through $5 very quickly.

---

## Rate Limits on Free Tier

New accounts without a deposit start at the **Free usage tier**:

- **Monthly spend limit:** $10 (your $5 credit fits within this)
- **Requests per minute:** ~5 RPM (varies by model)
- **Input tokens per minute:** ~20,000–50,000 ITPM depending on model
- **Output tokens per minute:** ~8,000–10,000 OTPM depending on model

To unlock higher limits, deposit $5 to reach **Build Tier 1** (50 RPM, $100/month spend limit). The deposit is separate from free credits — it requires a credit card.

---

## Credit Expiration & Billing

- **Free credits expire 14 days** after you click the "Claim" button
- **Purchased credits expire 1 year** from the purchase date
- All credit purchases are **non-refundable**
- After free credits are exhausted, you must **prepay** for additional credits (minimum $5 purchase) — there is no post-pay billing for standard accounts
- Credits are consumed by API calls, Workbench usage, and Claude Code

---

## Cost Optimization Features

Even with $5, you can use Anthropic's cost-saving features:

| Feature | Savings | Notes |
|---|---|---|
| **Batch API** | 50% off input and output | Asynchronous processing; results within 24 hours |
| **Prompt Caching (5-min)** | Cache reads at 10% of input price | Write cost is 1.25x base; great for repeated system prompts |
| **Prompt Caching (1-hour)** | Cache reads at 10% of input price | Write cost is 2x base; for longer sessions |
| **Shorter prompts** | Variable | Well-crafted prompts can reduce token usage by 50–80% |

---

## What Happens After Credits Run Out

1. API calls return errors — you will not be charged automatically
2. You must purchase more credits (minimum $5) with a credit card to continue
3. There is **no automatic billing** or surprise charges — Anthropic uses a prepaid model
4. Your API keys, projects, and settings remain intact

---

## Related Programs

- **Anthropic Startup Program** — $25,000 in credits for early-stage startups backed by partner VCs (Menlo Ventures, etc.). You cannot apply directly; your VC firm must be an Anthropic partner. Credits valid for 12 months
- **Claude via AWS Bedrock** — if you have AWS credits (e.g., via AWS Activate, up to $100K for startups), you can use Claude models through Amazon Bedrock without an Anthropic account
- **Claude via Google Cloud Vertex AI** — Google Cloud credits ($300 free trial or startup program credits up to $200K) can cover Claude API usage through Vertex AI
- **Claude via Microsoft Foundry** — Azure credits can also be used for Claude access
- **Anthropic Fellows Program** — 4-month research fellowship with ~$15K/month in compute funding and a stipend; aimed at AI safety researchers (next cohorts: May and July 2026)
- **Claude Student Ambassador Program** — campus program through Wasserman Next Gen; 10 weeks, 8 hours/week commitment

---

## Additional Tips

- **Free Claude.ai access** — if you just want to try Claude conversationally (not via API), [claude.ai](https://claude.ai) offers a free tier with daily usage limits. No credits needed
- **Workbench counts** — using the Workbench playground in the Console also consumes your API credits, not just direct API calls
- **Do not use a VPN** during signup — it can trigger security reviews and block credit claims
- **Plan before claiming** — since credits expire in 14 days, have your project and prompts ready before clicking "Claim"
- **Region restrictions** — the API is available in most countries including the US, Canada, UK, EU, Australia, Japan, and South Korea. Mainland China and certain other jurisdictions are excluded. Full list at [anthropic.com/supported-countries](https://www.anthropic.com/supported-countries)
- **Console URL has moved** — some older guides reference `console.anthropic.com`; the platform is now also accessible at `platform.claude.com`

---

**Sources:**
- [Anthropic API Pricing](https://platform.claude.com/docs/en/about-claude/pricing)
- [Anthropic Rate Limits](https://docs.anthropic.com/en/api/rate-limits)
- [Claude Phone Verification Help](https://support.claude.com/en/articles/8287232-why-do-i-need-to-verify-my-phone-number)
- [Claude API Free Credits Guide](https://www.aifreeapi.com/en/posts/claude-api-free-credits-guide)
- [Anthropic Supported Countries](https://www.anthropic.com/supported-countries)
- [Anthropic Startup Program](https://claude.com/programs/startups)
- [Anthropic Fellows Program 2026](https://alignment.anthropic.com/2025/anthropic-fellows-program-2026/)
- [Claude Plans & Pricing](https://claude.com/pricing)

## Submission 4: Anthropic AI for Science Program - Up to $20,000

- Category: Education & Student
- Source Link: https://www.anthropic.com/ai-for-science-program-rules
- Amount: Up to $20,000 in API credits over 6 months
- Difficulty: 4/5 (Hard)
- Expires At: N/A

### Description

Anthropic offers **up to $20,000 in API credits over 6 months** to researchers at recognized institutions working on high-impact scientific projects. The program emphasizes biology and life sciences but accepts applications from chemistry, medicine, environmental science, physics, computer science, and earth sciences. Applications are evaluated monthly on the first Monday of each month, based on scientific merit, potential impact, technical feasibility, and a mandatory biosecurity screening. No credit card is required — credits are awarded directly to your organizational API account.

---

## Registration (Step-by-Step)

1. Go to the [AI for Science application form](https://docs.google.com/forms/d/e/1FAIpQLSfwDGfVg2lHJ0cc0oF_ilEnjvr_r4_paYi7VLlr5cLNXASdvA/viewform?usp=header)
2. Describe your research project, your team credentials, and institutional affiliation
3. Explain how Claude API access would accelerate your work
4. Submit the form — applications are reviewed on a rolling monthly basis
5. If selected, you will be notified via email and must respond within **5 business days**
6. Sign any required compliance affidavits and liability releases within **14 days** of notification or forfeit the award
7. Credits are deposited into the organizational account you specified in the application

**Important:**
- You must be at least 18 years old (or the age of majority in your jurisdiction)
- You must be affiliated with a research institution — independent researchers without institutional backing do not qualify
- If employed, you may need employer/institution consent
- Anthropic employees, officers, directors, and their immediate families are ineligible
- Incomplete or non-compliant submissions may be disqualified without notice
- Anthropic does not send individual responses to unapproved submissions

---

## Evaluation Criteria

Applications are judged on four dimensions:

| Criterion | What Anthropic Looks For |
|---|---|
| **Scientific Merit** | Team credentials in their subject area and experience with AI, institutional affiliation |
| **Potential Impact** | Significance of the proposed research and its potential to advance the field |
| **Technical Feasibility** | Whether the project can realistically use Claude API to achieve its goals |
| **Biosecurity Screening** | Assessment to ensure the research does not raise concerns around accelerating harmful applications |

Selections are made on the **first Monday of each month** from all eligible entries received.

---

## Available API Models

Program participants receive access to Anthropic's **standard model suite** via the API. This currently includes:

| Model | Input (per MTok) | Output (per MTok) | Best For |
|---|---|---|---|
| **Claude Opus 4.5** | $5.00 | $25.00 | Flagship reasoning, complex analysis |
| **Claude Sonnet 4.5** | $3.00 | $15.00 | Balanced capability and cost |
| **Claude Haiku 4.5** | $1.00 | $5.00 | Fast, high-volume tasks |
| **Claude Opus 4** | $15.00 | $75.00 | Legacy, still available |
| **Claude Sonnet 4** | $3.00 | $15.00 | Legacy, still available |

**Key limitations:**
- Access is limited to the standard model suite — no non-standard or non-public models
- Credits apply **only to API usage** (via [console.anthropic.com](https://console.anthropic.com)), not to the Claude web app or Claude Pro subscription
- Fine-tuning is not covered by the program (available only through AWS Bedrock, in preview)

**Tip:** At Haiku 4.5 pricing ($1/$5 per MTok), $20,000 in credits translates to roughly **20 billion input tokens** or **4 billion output tokens**. At Opus 4.5 pricing ($5/$25), it translates to roughly **4 billion input tokens** or **800 million output tokens**. Choose models strategically based on your task complexity.

---

## Supported Research Fields

While the program has a strong emphasis on biology and life sciences, the following fields are explicitly supported:

- **Biology / Life Sciences** — understanding complex biological systems, genetic data analysis
- **Medicine / Healthcare** — drug discovery, especially for major global diseases
- **Chemistry**
- **Environmental Science** — including agricultural productivity
- **Physics**
- **Computer Science**
- **Earth Sciences**
- Other fields with significant scientific potential

---

## Multi-Researcher Teams

For research programs with multiple collaborators:

1. Set up a **single organizational account** to represent the entire program
2. Submit one application through the form on behalf of the team
3. If approved, credits are allocated to **one organization ID**
4. The team lead can then distribute API access among researchers within that account

This is the recommended approach — do not submit separate applications per team member.

---

## Country Restrictions

Applicants **cannot** be legal residents of:

- Belarus, China, Cuba, Iran, Myanmar, North Korea, Russia, Sudan, Syria
- Crimea, Donetsk People's Republic, or Luhansk People's Republic
- Any jurisdiction where receiving U.S. services is prohibited under applicable export control laws

---

## Usage Policy & Compliance

- All participants must comply with [Anthropic's Usage Policy](https://www.anthropic.com/policies/usage-policy)
- Anthropic's Trust & Safety team follows standard enforcement procedures for policy violations
- Red-teaming researchers receive **no special exemption** from usage policies
- Selected researchers grant Anthropic permission to use their names/likenesses for promotional purposes
- Researchers are responsible for any tax implications of receiving the credits
- Anthropic may cancel or modify the program at any time at its sole discretion

---

## Additional Tips

- **You do not need this program to use Claude API** — anyone can create a [Claude Console](https://console.anthropic.com) account and start using the API immediately with their own billing. The program simply provides free credits
- **Batch API discount** — Anthropic offers a 50% discount on asynchronous batch processing, which can stretch your credits significantly for large-scale data analysis
- **Prompt caching** — cached prompts cost only 10% of the original price, useful for repetitive analysis workflows common in research
- **No auto-billing** — program credits do not convert to a paid subscription; when credits run out, API access stops unless you add your own payment method
- **Anthropic Fellows Program** — a separate program for AI safety researchers, offering stipends and mentorship (not API credits)
- **Timeline expectation** — there is no guaranteed timeline for approval. If you are not selected, you will not receive a notification. You may reapply in subsequent months

---

**Sources:**
- [Anthropic AI for Science Program Rules](https://www.anthropic.com/ai-for-science-program-rules)
- [Introducing Anthropic's AI for Science Program](https://www.anthropic.com/news/ai-for-science-program)
- [Anthropic's AI for Science Program — Claude Help Center](https://support.claude.com/en/articles/11199177-anthropic-s-ai-for-science-program)
- [Anthropic Launches Program to Support Scientific Research — TechCrunch](https://techcrunch.com/2025/05/05/anthropic-launches-a-program-to-support-scientific-research/)
- [Claude API Pricing](https://platform.claude.com/docs/en/about-claude/pricing)
- [Claude Models Overview](https://platform.claude.com/docs/en/about-claude/models/overview)

## Submission 3: Alibaba Cloud Startup Catalyst - Up to $120,000

- Category: Startup Programs
- Source Link: https://www.alibabacloud.com/en/startup
- Amount: Up to $120,000 (lifetime maximum across Launcher $1K + Gold $20K + Platinum $40K + Diamond $59K)
- Difficulty: 4/5 (Hard)
- Expires At: 2026-03-31

### Description

Alibaba Cloud's **Startup Catalyst** (and its AI-focused variant, the **AI Catalyst Program**) offers startups a tiered credit system with a **$120,000 lifetime maximum** across four tiers: Launcher ($1,000), Gold ($20,000), Platinum ($40,000), and Diamond ($59,000). The program runs from **April 1, 2025 to March 31, 2026**. All startups begin at the Launcher tier and progress by submitting business proposals reviewed within 7 business days. Beyond credits, participants receive up to **2 billion free Model Studio tokens**, 1:1 AI expert office hours, a 12-month Alibaba Cloud Academy membership, and invitations to global Alibaba Cloud events.

---

## Registration (Step-by-Step)

1. Go to [alibabacloud.com/en/startup](https://www.alibabacloud.com/en/startup) (general program) or [alibabacloud.com/en/startup/ai](https://www.alibabacloud.com/en/startup/ai) (AI-focused variant)
2. **Create an Alibaba Cloud account** if you do not already have one -- email, phone verification, and payment method required
3. Click **"Apply Now"** and fill in the online application form
4. Prepare and submit a **Business Proposal** that includes:
   - Plans for future cloud credit usage
   - Monthly estimated cloud credit consumption
   - Growth plans and projected scaling
5. You may receive an email requesting **additional information** -- respond promptly
6. Wait for review -- Alibaba Cloud provides feedback or a response **within 4-7 business days**
7. If feedback is given, **incorporate it and resubmit** the proposal for final review
8. Upon approval, you receive Launcher tier credits ($1,000) in your Alibaba Cloud account
9. To advance to higher tiers (Gold, Platinum, Diamond), submit additional proposals demonstrating startup funding and growth

**Important:**
- Every startup starts at the **Launcher tier** regardless of size
- Tier progression requires meeting Alibaba Cloud's internal criteria (startup funding, projected growth, cloud usage)
- Business proposals may go through multiple review rounds
- Credits are non-transferable and cannot be shared with end customers
- Using multiple accounts to exceed the $120,000 lifetime cap may result in forfeiture of excess credits

---

## Credit Tiers & Validity

| Tier | Credit Amount | Validity Period |
|---|---|---|
| **Launcher** | $1,000 | 3 months from redemption |
| **Gold** | $20,000 | 6 months from redemption |
| **Platinum** | $40,000 | 9 months from redemption |
| **Diamond** | $59,000 | 12 months from redemption |

- **Lifetime maximum:** $120,000 across all tiers combined ($1,000 + $20,000 + $40,000 + $59,000 = $120,000)
- Credits at each tier have their own validity window starting from the date of redemption
- The program campaign period runs **April 1, 2025 -- March 31, 2026**, but individual credit validity is based on redemption date

---

## What Credits Cover (and Exclusions)

Credits come in two forms:

**1. Service credits** -- applicable to most Alibaba Cloud products including:
- Compute (ECS, Function Compute, Container Service)
- Databases (ApsaraDB RDS, PolarDB, Redis)
- AI/ML (Model Studio, Machine Learning Platform)
- Storage (OSS, NAS)
- Networking (VPC, SLB, Elastic IP)
- Big Data (MaxCompute, DataWorks)
- Security services

**2. Business Support credits** -- for Alibaba Cloud technical support plans

**Excluded from credits:**
- Domain name purchases
- CDN (Content Delivery Network)
- Alibaba Cloud Academy courses
- Alibaba Cloud Marketplace purchases
- Prepaid products and services

---

## AI Catalyst Program Benefits

The AI-focused variant of the program adds AI-specific perks on top of the standard credit tiers:

| Benefit | Details |
|---|---|
| **Model Studio tokens** | Up to 2 billion free tokens for Qwen-series models |
| **1:1 Office Hours** | Sessions with AI specialists covering products, solutions, and infrastructure |
| **Cloud Academy** | 12-month membership for cloud and AI learning with certification |
| **Global events** | Invitations to Alibaba Cloud events and Apsara Conference |
| **AI Alliance network** | Connections to clients, partners, investors, and incubators |
| **Co-marketing** | Joint marketing initiatives to expand customer reach |
| **POC coupons** | Proof-of-concept support coupons |
| **Tech support** | 3-month Global Technology Services package for the price of 1 month |

---

## Available AI Models (Model Studio)

Participants get access to Alibaba Cloud's Model Studio platform with Qwen-series models:

| Model | Use Case |
|---|---|
| **Qwen-Max (MoE)** | Flagship reasoning model |
| **Qwen-Plus** | Balanced performance and cost |
| **Qwen-Turbo** | Fast, cost-effective inference |
| **Qwen-VL** | Vision-language understanding |
| **Qwen3-Coder-Plus** | Code generation |
| **Qwen-MT-Plus** | Translation (92 languages) |
| **Wan series** | Image and video generation |

**Note:** Model Studio free tokens cover real-time inference only (not batch calls or context cache). The token quota is shared across an Alibaba Cloud account and its RAM sub-users.

---

## Eligibility & Requirements

- Must have a **valid Alibaba Cloud international account**
- Targeted at **early-stage startups** with innovative solutions
- No specific industry restriction -- AI, SaaS, fintech, e-commerce, and other verticals are eligible
- Available **globally** (Alibaba Cloud international regions)
- Startup funding stage and growth trajectory are evaluated during tier progression
- Agents, resellers, and distributor sub-accounts may face restrictions
- Alibaba Cloud reserves the right to accept or reject applications at its discretion

---

## Restrictions & Gotchas

- **Tier progression is not automatic** -- you must submit new proposals and meet requirements for each upgrade
- **Credits expire per tier** -- unused Launcher credits expire after 3 months even if you have not yet applied for Gold
- **Non-transferable** -- credits cannot be given to customers or other accounts
- **Multiple accounts prohibited** -- exceeding $120,000 via multiple accounts results in forfeiture
- **Alibaba Cloud can terminate participation** for security risks, liability concerns, or suspected fraud
- **Auto-billing risk** -- once credits are exhausted, PAYG services will bill your payment method at standard rates
- **Excluded services** are billed normally even during the program (domain, CDN, Marketplace, Academy, prepaid)
- **Proposal resubmission** -- Alibaba Cloud may require revisions before approval, adding time to the process

---

## Additional Tips

- **Apply to the AI Catalyst variant** if you are building with AI -- you get the same credit tiers plus 2 billion Model Studio tokens and AI-specific mentoring
- **Start with Launcher immediately** -- the $1,000 is granted quickly and lets you prototype while preparing a Gold-tier proposal
- **Set up billing alerts** in the Alibaba Cloud console to avoid surprise charges when credits expire
- **Enable "Free Quota Only"** toggle in Model Studio to prevent charges after token quota is exhausted
- **Back up data before credit expiry** -- resources may be deleted if you do not upgrade to paid plans
- **Alibaba Cloud Free Trial** is a separate program ($1,700 for individuals, $8,500 for enterprises) that can be combined if you are a new user
- **Market immersion trips** -- the program occasionally organizes trips for startups to explore new markets and meet partners
- **Referral and co-marketing** opportunities can provide additional exposure for your startup

---

**Sources:**
- [Alibaba Cloud Startup Catalyst Program](https://www.alibabacloud.com/en/startup)
- [Alibaba Cloud AI Catalyst Program](https://www.alibabacloud.com/en/startup/ai)
- [Startup Cloud Credits Terms & Conditions](https://www.alibabacloud.com/en/startup/terms-and-conditions)
- [KrASIA - Alibaba Cloud Unveils Catalyst Program](https://kr-asia.com/alibaba-cloud-unveils-catalyst-program-to-support-international-startups)
- [Holori - Free Cloud Credits Guide](https://holori.com/maximize-free-cloud-credit/)
- [CloudCredits.io - Alibaba Cloud](https://cloudcredits.io/providers/alibabacloud/)
- [Alibaba Cloud Model Studio](https://www.alibabacloud.com/en/product/modelstudio)

## Submission 2: Alibaba Cloud Model Studio - Free Token Quota

- Category: Cloud Providers
- Source Link: https://www.alibabacloud.com/help/en/model-studio/new-free-quota
- Amount: 1,000,000 tokens per model (text/code/vision/reasoning), 100 images (image gen), 50 seconds (video gen), 2,000-10,000 characters (TTS)
- Difficulty: 1/5 (Instant)
- Expires At: N/A

### Description

First-time activators of **Alibaba Cloud Model Studio** (Singapore region) receive a **free token quota for every available model** — typically **1 million tokens per model**, valid for **90 days**. The quota covers the full Qwen model family (text, vision, code, reasoning, multimodal, image generation, video generation, and TTS). Quotas are **per-model and not shared** between model versions (e.g., `qwen-max` and `qwen-max-latest` have separate 1M allocations). No credit card is required for activation. The API is **OpenAI-compatible** — just swap the base URL and API key.

---

## How to Activate

1. Create an **Alibaba Cloud International** account at [alibabacloud.com](https://www.alibabacloud.com)
2. Go to the **Model Studio console** (Singapore region): [Model Studio Console](https://dashscope.console.alibabacloud.com/)
3. Read and accept the **Terms of Service**
4. Model Studio activates automatically — free quotas are granted immediately
5. Generate an **API key** (called DashScope API key) from the console
6. Start making API calls using the OpenAI-compatible endpoint

**No credit card required.** The free quota begins counting from the moment you activate Model Studio.

---

## Free Quota Per Model

### Text Generation Models

| Model | Context Window | Free Quota | Input Price (after free) | Output Price (after free) |
|---|---|---|---|---|
| **qwen3-max** | 262K tokens | 1M tokens / 90 days | $1.20/M | $6.00/M |
| **qwen-plus** | 1M tokens | 1M tokens / 90 days | $0.40/M | $1.20/M |
| **qwen-flash** | 1M tokens | 1M tokens / 90 days | $0.05/M | $0.40/M |
| **qwen-turbo** | — | 1M tokens / 90 days | $0.05/M | $0.20/M |
| **qwq-plus** (reasoning) | 131K tokens | 1M tokens / 90 days | $0.80/M | $2.40/M |

### Code Models

| Model | Free Quota | Input Price | Output Price |
|---|---|---|---|
| **qwen3-coder-plus** | 1M tokens / 90 days | $1.00/M | $5.00/M |
| **qwen3-coder-flash** | 1M tokens / 90 days | $0.30/M | $1.50/M |

### Vision & Multimodal Models

| Model | Context Window | Free Quota | Input Price | Output Price |
|---|---|---|---|---|
| **qwen3-vl-plus** | 262K tokens | 1M tokens / 90 days | $0.20/M | $1.60/M |
| **qwen3-vl-flash** | 262K tokens | 1M tokens / 90 days | $0.05/M | $0.40/M |
| **qvq-max** (visual reasoning) | 131K tokens | 1M tokens / 90 days | $1.20/M | $4.80/M |
| **qwen3-omni-flash** | 65K tokens | 1M tokens / 90 days | $0.43/M (text) | $1.66/M (text) |

### Open-Source Hosted Models

| Model | Free Quota | Input Price | Output Price |
|---|---|---|---|
| **qwen3-235b-a22b** | 1M tokens / 90 days | $0.70/M | $2.80/M |
| **qwen3-32b** | 1M tokens / 90 days | $0.16/M | $0.64/M |
| **qwen3-8b** | 1M tokens / 90 days | $0.18/M | $0.70/M |
| **qwen2.5-72b-instruct** | 1M tokens / 90 days | $1.40/M | $5.60/M |
| **qwen2.5-32b-instruct** | 1M tokens / 90 days | $0.70/M | $2.80/M |
| **qwen2.5-14b-instruct** | 1M tokens / 90 days | $0.35/M | $1.40/M |
| **qwen2.5-7b-instruct** | 1M tokens / 90 days | $0.175/M | $0.70/M |

### Translation

| Model | Free Quota | Input Price | Output Price |
|---|---|---|---|
| **qwen-mt-plus** | 1M tokens / 90 days | $2.46/M | $7.37/M |

### Image Generation

| Model | Free Quota | Price (after free) |
|---|---|---|
| **qwen-image-max** | 100 images / 90 days | $0.075/image |
| **qwen-image-edit-max** | 100 images / 90 days | $0.075/image |

### Video Generation

| Model | Free Quota | Price (after free) |
|---|---|---|
| **wan2.6-t2v** (text-to-video) | 50 seconds / 90 days | $0.10–$0.15/sec |
| **wan2.6-i2v** (image-to-video) | 50 seconds / 90 days | $0.10–$0.15/sec |

### Text-to-Speech

| Model | Free Quota | Price (after free) |
|---|---|---|
| **qwen3-tts-flash** | 2,000–10,000 chars / 90 days | $0.01/1K chars |

---

## API Usage (OpenAI-Compatible)

The API is fully compatible with the OpenAI SDK. Just change the base URL and API key.

**Base URL (Singapore):** `https://dashscope-intl.aliyuncs.com/compatible-mode/v1`

**Python example:**
```python
from openai import OpenAI
import os

client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

response = client.chat.completions.create(
    model="qwen-plus",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)
```

**cURL example:**
```bash
curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \
  -H "Authorization: Bearer $DASHSCOPE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "qwen-plus", "messages": [{"role": "user", "content": "Hello!"}]}'
```

The newer **Responses API** uses a different base URL:
`https://dashscope-intl.aliyuncs.com/api/v2/apps/protocols/compatible-mode/v1`

---

## Region Restrictions

- **Singapore (International Edition):** Free quota available. This is the only region with free tokens.
- **Mainland China (Beijing):** No free quota. Third-party models (DeepSeek, Kimi) are available but only in this region.
- **US (Virginia):** Available but typically no free quota.
- Each region requires its own separate API key.

---

## Important Notes & Tips

- **Quota is shared** across your main Alibaba Cloud account and all RAM (sub) users. If the main account uses 300K tokens of qwen-max's free quota, RAM users share the remaining 700K.
- **"Free Quota Only" toggle** — enable this in the console to prevent accidental charges after the free quota runs out. When enabled, API calls return an `AllocationQuota.FreeTierOnly` error instead of billing you.
- **No depletion notification** — there is currently no alert when your free quota is about to run out. Check remaining tokens manually in the console.
- **Console updates hourly** — the remaining quota shown in the console is not real-time; it refreshes approximately every hour.
- **Batch calls excluded** — OpenAI-compatible Batch API calls do not consume the free quota (they are billed separately at 50% discount).
- **Context cache fees excluded** — if you use context caching, the cache storage fees are not covered by the free quota.
- **Snapshot versions get separate quotas** — e.g., `qwen3-max`, `qwen3-max-2025-09-23`, and `qwen3-max-2026-01-23` each get their own 1M free tokens.
- **Savings plans** — after the free tier, you can pre-purchase token bundles ($10–$5,000) for discounted rates.
- **90-day validity** applies to activations after September 8, 2025. Earlier activations may have shorter validity periods (30 days).

---

**Sources:**
- [Free Quota for New Users](https://www.alibabacloud.com/help/en/model-studio/new-free-quota)
- [Model Studio Models](https://www.alibabacloud.com/help/en/model-studio/models)
- [Model Pricing](https://www.alibabacloud.com/help/en/model-studio/billing-for-model-studio)
- [OpenAI-Compatible API](https://www.alibabacloud.com/help/en/model-studio/compatibility-of-openai-with-dashscope)
- [Free Quota Validity Adjustment Notice](https://www.alibabacloud.com/help/en/model-studio/new-free-quota-validity-adjustment)
- [What is Model Studio](https://www.alibabacloud.com/help/en/model-studio/what-is-model-studio)

## Submission 1: Alibaba Cloud Free Trial - Up to $1,700

- Category: Cloud Providers
- Source Link: https://www.alibabacloud.com/en/free
- Amount: Up to $1,700 (individuals) or $8,500 (enterprises)
- Difficulty: 2/5 (Easy)
- Expires At: N/A

### Description

New Alibaba Cloud users get **up to $1,700 in free trial credits** (individuals) or **$8,500** (verified enterprises) across **80+ products** spanning compute, storage, networking, security, big data, and AI. The ECS (Elastic Compute Service) free trial is valid for **up to 12 months**; other PAYG product credits expire after **60 days**. Credit card required at signup. Additionally, **20+ always-free products** remain available indefinitely with limited capacity.

---

## Registration (Step-by-Step)

1. Go to [alibabacloud.com/en/free](https://www.alibabacloud.com/en/free) and click **"Free Trial"**
2. Select account type: **Individual** or **Enterprise**
3. Enter your **email address**, click Send, and enter the verification code from your inbox
4. Accept the Alibaba Cloud Terms of Service and click **Register**
5. **Phone verification** -- enter your mobile number and submit the SMS verification code
6. **Payment method** -- add a credit card. Alibaba Cloud places a small temporary hold (under $1, refunded) and sends a 6-digit verification code in the transaction record. Enter either the code or the charged amount to verify
7. For **Enterprise** accounts: submit company name, registration number, registered address, country, and a photo of your company certificate. Review takes 1-2 business days
8. Once verified, go to the Free Trial page and activate the products you want

**Important:**
- Only accounts created **within the past 1 year** are eligible
- One free trial per person or organization
- Individual and Enterprise trials are mutually exclusive -- you cannot switch from Individual to Enterprise
- Sub-accounts of agents, resellers, and distributors are not eligible
- ECS trial packages are subscription-type -- no auto-billing after expiration
- Other PAYG credits will auto-charge at standard rates if exceeded

---

## Free Trial Credit Structure

| Tier | Credit Value | ECS Duration | Other Products |
|---|---|---|---|
| **Individual** | Up to $1,700 | Up to 12 months | 60 days PAYG credits |
| **Enterprise** (verified) | Up to $8,500 | Up to 12 months (higher specs) | 60 days PAYG credits |

---

## Key Free Trial Products

| Category | Products |
|---|---|
| **Compute** | ECS (Elastic Compute Service) c7/g7 4C16G instances, Function Compute |
| **Database** | ApsaraDB RDS, ApsaraDB for PolarDB, ApsaraDB for Redis |
| **Storage** | Object Storage Service (OSS) |
| **Networking** | Virtual Private Cloud (VPC), Elastic IP, Server Load Balancer |
| **Security** | Cloud Firewall, Anti-DDoS Basic, Security Center (7-day trial) |
| **Big Data** | MaxCompute, DataWorks |
| **AI/ML** | Machine Learning Platform for AI, Model Studio |
| **Containers** | Container Service for Kubernetes (ACK) |

---

## AI Services & Model Studio

Alibaba Cloud Model Studio provides access to Qwen-series LLMs with a **free inference quota for new users** (valid for **90 days**, Singapore region only):

| Model | Free Quota (example) | Notes |
|---|---|---|
| **Qwen-Max / Qwen3-Max** | ~1,000,000 tokens | Flagship reasoning model |
| **Qwen-Plus / Qwen3-Plus** | Free quota available | Balanced performance |
| **Qwen-Turbo** | Free quota available | Fast, cost-effective |
| **Qwen-VL / Qwen3-VL-Plus** | Free quota available | Vision-language models |
| **Qwen3-Coder-Plus** | Free quota available | Code generation |
| **Qwen-MT-Plus** | Free quota available | Translation (92 languages) |
| **Wan series** | Free quota available | Image/video generation |

**Model Studio notes:**
- Free quota covers **real-time inference only** (not batch calls or context cache)
- Available **only in the Singapore region**
- Enable **"Free Quota Only"** toggle in the console to avoid charges after quota is exhausted
- An Alibaba Cloud account and its RAM users share the same free quota
- Quota usage in the console updates hourly, not in real time

**GenAI Empowerment Program:** Startups and developers building with Qwen models can apply for additional free cloud credits, training workshops, and co-marketing opportunities.

---

## Always-Free Products (No Expiry)

These 20+ services remain free indefinitely with limited capacity:

| Service | Description |
|---|---|
| **CloudMonitor** | Infrastructure monitoring and alerting |
| **Auto Scaling** | Automatic ECS instance scaling |
| **Container Registry** | Container image hosting and lifecycle management |
| **RAM (Resource Access Management)** | Identity and access management |
| **Key Management Service (KMS)** | Cryptography and key management |
| **Anti-DDoS Basic** | Basic DDoS protection for ECS instances |
| **DataWorks** | Big data development and governance |
| **Database Autonomy Service (DAS)** | Self-healing database monitoring |
| **Data Management (DMS)** | Database R&D service platform |
| **Operation Orchestration Service (OOS)** | Automated O&M tasks |
| **Resource Orchestration Service** | Cloud resource deployment automation |
| **Web App Service** | PaaS for web, mobile, and API apps |

---

## Available Regions

Free trial resources can be deployed across these regions:

| Area | Regions |
|---|---|
| **China** | Hangzhou, Shanghai, Nanjing, Qingdao, Beijing, Zhangjiakou, Hohhot, Ulanqab, Shenzhen, Heyuan, Guangzhou, Chengdu |
| **Asia Pacific** | Hong Kong, Singapore, Tokyo, Seoul, Kuala Lumpur, Jakarta, Manila, Bangkok |
| **Americas** | US (Silicon Valley), US (Virginia) |
| **Europe & Middle East** | Frankfurt, London, Dubai, Riyadh |

**Note:** Model Studio AI free quota is only available in the **Singapore** region. No specific country exclusions are published, but eligibility depends on local regulations and Alibaba Cloud international site terms.

---

## Restrictions & Limitations

- **Credit card required** -- autopay activates after trial ends for PAYG products
- **OS lock-in** -- you are restricted to the OS (Linux or Windows) selected at ECS instance creation
- **Do not modify ECS specs** -- changing instance type, quantity, disk type/size, or bandwidth during trial may change the price from $0
- **No accumulation** -- free usage does not roll over month to month
- **Prohibited activities** -- bitcoin mining, fraud, illegal activities, or terms violations will result in disqualification
- **Data deletion** -- if you do not upgrade after trial ends, Alibaba Cloud may delete your data

---

## Tips to Maximize Value

1. **Choose Enterprise verification if eligible** -- you get 5x the credit value ($8,500 vs $1,700) and higher ECS specifications
2. **Activate Model Studio early** -- the 90-day AI quota clock starts at activation, not at first API call. Enable "Free Quota Only" toggle immediately
3. **Use a credit card, not PayPal** -- PayPal cannot pay for PAYG services, limiting flexibility
4. **Monitor usage in the console** -- set up CloudMonitor alerts (always free) to track consumption and avoid surprise charges
5. **Back up data before trial ends** -- especially ECS instances and OSS buckets. Resources may be deleted after expiration
6. **ECS renewal discount** -- after receiving the free trial, you automatically get a minimum 40% discount for ECS renewals and new purchases
7. **Referral program** -- refer others to the free tier to earn coupon credits
8. **Cancel PAYG services before 60 days** if you do not plan to continue
9. **Start with always-free products** for long-term workloads (CloudMonitor, RAM, Auto Scaling) and use trial credits for premium services
10. **Check the GenAI Empowerment Program** if you are a startup building with Qwen models -- additional credits and support available

---

**Sources:**
- [Alibaba Cloud Free Trial Page](https://www.alibabacloud.com/en/free)
- [Alibaba Cloud Free Trial Blog](https://www.alibabacloud.com/blog/alibaba-cloud-free-trial-how-to-sign-up-and-get-started_595840)
- [Account Registration Guide](https://www.alibabacloud.com/help/en/account/step-1-register-an-alibaba-cloud-account)
- [Corporate Identity Verification](https://www.alibabacloud.com/help/en/account/user-guide/corporate-identity-verification)
- [Model Studio Free Quota](https://www.alibabacloud.com/help/en/model-studio/new-free-quota)
- [Model Studio Pricing](https://www.alibabacloud.com/help/en/model-studio/billing-for-model-studio)
- [ECS Free Trial Introduction](https://www.alibabacloud.com/blog/an-introduction-to-ecs-free-trial-offer_598861)
- [OSS Free Quota for New Users](https://www.alibabacloud.com/help/en/oss/free-quota-for-new-users)
- [Alibaba Cloud Free Tier Overview](https://albertoroura.com/alibaba-cloud-free-tier-and-free-services/)