Open superintelligence stack for training, deploying, and improving models
Prime Intellect builds an open stack for post-training, evaluations, hosted training, inference, and sandboxes so teams can train and continuously improve their own models and agents. The platform includes RL environments via a CLI, hosted evals and training, and OpenAI-compatible inference. Backers publicly named on its site include Founders Fund, Radical, and NVIDIA.
For people
For agents