Prime Intellect Fast Compute Grants ($500-$100K)
Source: https://www.primeintellect.ai/
Description
Create account to comment on specific lines or Sign in
+ 1 Prime Intellect runs Fast Compute Grants — a rolling, low-bureaucracy grant program awarding $500-$100,000 in compute credits on its GPU platform to researchers and builders working on open-source / decentralized AI. The program was launched in 2024 to back "ambitious open-source AI research into novel models, architectures, techniques, and paradigms needed to make open and decentralized AGI a reality." Decisions are returned in 5-10 days and applications are made by emailing a pitch to contact@primeintellect.ai. There are no formal eligibility hoops (no VC backing, no .edu requirement, no nationality restriction) — but the technical bar is explicitly described as high.
No comments on this line yet.
+ 2
No comments on this line yet.
+ 3 Credits are spent on Prime Intellect Compute, an aggregator that brokers GPUs from 50+ underlying providers (effectively a meta-cloud). The platform exposes both on-demand instances (single GPUs up to 256-GPU clusters) and reserved clusters, with hourly billing and the option to resell idle reserved capacity back to the spot market.
No comments on this line yet.
+ 4
No comments on this line yet.
+
5
No comments on this line yet.
+ 6
No comments on this line yet.
+ 8
No comments on this line yet.
+ 9 Good fit:
No comments on this line yet.
+ 10 • Independent researchers, PhD students, postdocs working on open-source AI
No comments on this line yet.
+ 11 • Small teams doing distributed / decentralized training research (federated learning, swarm training, model merging, etc.)
No comments on this line yet.
+ 12 • Builders shipping open-weight models, novel architectures, RL post-training experiments, evals or environments
No comments on this line yet.
+ 13 • Hackers who already have a concrete experiment plan and just need GPUs (not "I want to learn ML")
No comments on this line yet.
+ 14
No comments on this line yet.
+ 15 Bad fit:
No comments on this line yet.
+ 16 • Closed-source / proprietary product work — the program is explicitly oriented toward open AI
No comments on this line yet.
+ 17 • People without a clear technical proposal — "the bar for quality is high"
No comments on this line yet.
+ 18 • Anyone needing inference for a SaaS product (use the standard paid platform or one of the cheap free tiers like Groq / Cerebras / Z.AI)
No comments on this line yet.
+ 19
No comments on this line yet.
+
20
No comments on this line yet.
+ 21
No comments on this line yet.
+ 23
No comments on this line yet.
+ 24 1. Draft a one-page pitch. No formal template, but cover: who you are, what you want to build/research, why it matters for open AI, GPU type + hours you need, and a rough timeline. Link any prior work (GitHub, papers, blog posts).
No comments on this line yet.
+
25
2. Email the pitch to contact@primeintellect.ai with a subject line like Fast Compute Grants Application — <your project>.
No comments on this line yet.
+ 26 3. Wait 5-10 days for a decision. Approved grants come with credits provisioned to a Prime Intellect account.
No comments on this line yet.
+ 27 4. Create an account at app.primeintellect.ai (Google or email signup) — you can do this before applying so the credits attach to an existing workspace.
No comments on this line yet.
+ 28 5. Spend the credits via the dashboard: launch on-demand pods, reserve clusters, or use the platform's RL / training stack.
No comments on this line yet.
+ 29
No comments on this line yet.
+ 30 Important:
No comments on this line yet.
+ 31 • This is email-based intake — there is no public web form. Do not wait for a fancy application portal.
No comments on this line yet.
+ 32 • "Anyone from anywhere can apply" — no US/EU residency requirement, no need to be at a university or company.
No comments on this line yet.
+ 33 • The credit amount you get is sized to the proposal; don't ask for $100K to fine-tune a 7B on a small dataset.
No comments on this line yet.
+ 34
No comments on this line yet.
+
35
No comments on this line yet.
+ 36
No comments on this line yet.
+ 38
No comments on this line yet.
+ 39 Credits are spent on Prime Intellect Compute, which brokers GPUs across 50+ providers. As of May 2026 the on-demand catalog includes:
No comments on this line yet.
+ 40
No comments on this line yet.
+ 41 GPUVRAMApprox. on-demand price
No comments on this line yet.
+ 42 NVIDIA B300288 GB~$4.99 / hr
No comments on this line yet.
+ 43 NVIDIA B200192 GB~$3.49 / hr
No comments on this line yet.
+ 44 NVIDIA H20080 GB (HBM3e)~$0.47-$1.99 / hr
No comments on this line yet.
+ 45 NVIDIA H10080 GB~$0.94 / hr (spot)
No comments on this line yet.
+ 46 NVIDIA GH200, A100, A40 and other professional/legacy SKUsvariesvaries
No comments on this line yet.
+ 47
No comments on this line yet.
+ 48 Cluster sizes scale from 1 GPU up to 256-GPU on-demand clusters; larger reserved clusters are quoted within 24h. Pricing varies because the underlying providers compete on price — you can pick the cheapest provider for a given SKU at booking time.
No comments on this line yet.
+ 49
No comments on this line yet.
+ 50 A $500 grant is enough for ~250-1,000 H100 hours. A $100K grant enables a serious training run on a multi-node H100/H200 cluster.
No comments on this line yet.
+ 51
No comments on this line yet.
+ 52 See the live catalog at app.primeintellect.ai/dashboard/create-cluster.
No comments on this line yet.
+ 53
No comments on this line yet.
+
54
No comments on this line yet.
+ 55
No comments on this line yet.
+ 57
No comments on this line yet.
+ 58 Prime Intellect is the company behind several open distributed-training projects (INTELLECT-1, OpenDiloco, decentralized RL post-training). Proposals that align with their thesis tend to be received better:
No comments on this line yet.
+ 59
No comments on this line yet.
+ 60 • Distributed / decentralized training — federated learning, low-communication optimizers, swarm training, model merging
No comments on this line yet.
+ 61 • Open-weight model training and fine-tuning — especially novel architectures or training recipes
No comments on this line yet.
+ 62 • RL post-training and agent research — reasoning models, tool-use agents, multi-turn environments
No comments on this line yet.
+ 63 • Open evals, environments, and benchmarks — anything that becomes a public artifact
No comments on this line yet.
+ 64 • Inference research — speculative decoding, quantization, serving optimizations on open models
No comments on this line yet.
+ 65
No comments on this line yet.
+ 66 Closed-source product work, generic API consumption, and "learning projects" are less likely to be funded.
No comments on this line yet.
+ 67
No comments on this line yet.
+
68
No comments on this line yet.
+ 69
No comments on this line yet.
+ 71
No comments on this line yet.
+ 72 • Bar is quality, not paperwork. Selection is discretionary and based on the technical merit of the proposal and the credibility of the applicant. There is no published acceptance rate.
No comments on this line yet.
+ 73 • Open-by-default expectation. While not always a hard requirement, grantees are encouraged (and frequently expected) to publish their work — code, models, write-ups — so the broader open-AI community benefits.
No comments on this line yet.
+ 74 • Promotion in exchange for visibility. Selected projects are amplified through Prime Intellect's channels and partner network.
No comments on this line yet.
+ 75 • Credits are platform-locked. They can only be spent on Prime Intellect Compute — not transferred to other providers or cashed out.
No comments on this line yet.
+ 76 • No published expiry, but treat the credits as use-it-or-lose-it within a few months; clarify the window when accepting.
No comments on this line yet.
+ 77
No comments on this line yet.
+
78
No comments on this line yet.
+ 79
No comments on this line yet.
+ 81
No comments on this line yet.
+ 82 • $100 free credits for Akash GPU integration users — separate small promo for users trying the Akash decentralized GPU pool through Prime Intellect.
No comments on this line yet.
+ 83 • Reserved-cluster spot resale — once you hold a reserved cluster you can resell idle hours back to the spot market, effectively reducing your net burn.
No comments on this line yet.
+ 84 • ICLR / conference promos — Prime Intellect periodically announces conference-tied free-compute pushes (ICLR 2024, NeurIPS, etc.). Worth checking their X/Twitter (@PrimeIntellect) before relevant deadlines.
No comments on this line yet.
+ 85
No comments on this line yet.
+
86
No comments on this line yet.
+ 87
No comments on this line yet.
+ 89
No comments on this line yet.
+ 90 • Apply with a concrete experiment, not an idea. "I want to train an MoE" is weaker than "I want to reproduce X paper at Y scale on Z dataset, expected cost ~600 H100-hours, here is the training script."
No comments on this line yet.
+ 91 • Show prior work. A GitHub repo or short technical blog post dramatically improves your odds — it removes the reviewer's risk that you can't actually use the GPUs.
No comments on this line yet.
+ 92 • Cite Prime Intellect's own research if relevant (INTELLECT-1, OpenDiloco, PRIME-RL). Alignment with their roadmap is a tailwind.
No comments on this line yet.
+ 93 • Don't pad the ask. Asking for $100K when you really need $5K is a red flag; asking for what you actually need is a positive signal.
No comments on this line yet.
+ 94 • If rejected, ship something small first on the cheap on-demand tier (~$0.47/hr H200 spot) and re-apply with results.
No comments on this line yet.
+ 95 • Cheaper alternatives for small experiments: free tiers from Groq, Cerebras, SambaNova, Together AI, or the always-free Z.AI GLM-Flash models. Use Fast Compute Grants for training, not inference exploration.
No comments on this line yet.
+ 96
No comments on this line yet.
+
97
No comments on this line yet.
+ 98
No comments on this line yet.
+ 99 Sources:
No comments on this line yet.
+ 100 • Prime Intellect — homepage
No comments on this line yet.
+ 101 • Prime Intellect on X — Fast Compute Grants launch
No comments on this line yet.
+ 102 • Prime Intellect on X — Open and decentralized AGI grant call
No comments on this line yet.
+ 103 • Prime Intellect — LinkedIn announcement
No comments on this line yet.
+ 104 • Prime Intellect Blog — Compute exchange
No comments on this line yet.
+ 105 • Prime Intellect Blog — $15M fundraise
No comments on this line yet.
+ 106 • GPU-Grants list (community-maintained)
No comments on this line yet.
+ 107 • Prime Intellect docs
No comments on this line yet.