About

Prime Intellect is a compute and infrastructure platform for training, evaluating, and deploying agentic models. The stack covers Lab for hosted RL training and deployment, on-demand GPU access from 1 to 256 GPUs, and Liquid Reserved Clusters running 8 to 5000+ GPUs across 50+ providers. Sandboxes handle secure code execution for large-scale RL.

The team also publishes its own research models and open source. INTELLECT-3 is a 100B+ parameter Mixture-of-Experts model trained on the in-house RL stack (Nov 2025). Intellect 2 was a 32B model trained through globally distributed reinforcement learning (May 2025). The Environments Hub hosts open-source RL environments, alongside Verifiers and Prime-RL frameworks.

Funding

Backed by Founders Fund, with named angels including Andrej Karpathy, Dylan Patel, Clem Delangue, and Tri Dao.