Prime Intellect operates an open-source RL stack - the Environments Hub (2,500+ community RL environments), the Verifiers library and prime-rl training framework, plus hosted RL post-training (Lab), evals, inference and on-demand GPU compute. It positions itself as the open alternative to closed big-lab RL tooling and also trains open models (INTELLECT series).
How we verified this
Re-verified Prime Intellect against primary and third-party sources. Company identity confirmed correct - this is the open-source RL/environments company ('GitHub for RL environments', Environments Hub, Verifiers, prime-rl, INTELLECT models), matching the directory note as the open-source backbone of the category. Founders confirmed: Vincent Weisser (CEO) and Johannes Hagemann (CTO), founded 2023; Will Brown is a research lead (Verifiers creator), not a founder. Headcount sanity-checks out at 11-50 (LinkedIn public snippet; Sacra 23 FTE) - not 200+. Key correction: the 'Endless Terminals' paper (arXiv:2601.16443) was misattributed to Prime Intellect; its actual authors are Kanishk Gandhi, Shivam Garg, Noah Goodman, Dimitris Papailiopoulos with no Prime Intellect affiliation, so it was removed. INTELLECT-2 and INTELLECT-3 are genuine Prime Intellect publications (INTELLECT-3 arXiv:2512.16144 authored by 'Prime Intellect Team'). Funding: $5.5M seed (Apr 2024, Distributed Global + CoinFund) and $15M (Feb 2025, Founders Fund lead) are independently confirmed; the $70.4M total and ~$49.9M Series B (Dec 2025) rest only on the Tracxn aggregator with no primary announcement found, so they were flagged as single-source while kept at 'reported'. Valuation unknown. NVIDIA relationship upgraded to 'verified' (confirmed by NVIDIA's own newsroom - Prime Intellect integrates NeMo Gym) with frontier_lab_tie true, but reframed as an integration rather than a customer; all other named customers (Zapier, Ramp, Arcee, Character AI, Browserbase, Groq) remain self-claimed (vendor-site only). No SOC 2, certifications, or trust/security page found - all left unknown. Overall confidence lowered from high to medium due to the misattribution and single-source Series B.