Prime Intellect

How prime-rl transfers a 1.6 TB GLM-5.2 policy into DPEP=32 vLLM inference, overlaps online quantization, and reduces end-to-end synchronization from 86.1 seconds to 3.9 seconds.

We found models circumventing offline evaluation restrictions through inference API remote-fetch capabilities and coordinated fixes across affected frameworks.

Compare 153 autonomous runs from 18 frontier models on the nanoGPT optimizer speedrun.

We ran 153 autonomous runs across 18 frontier models on the nanoGPT optimizer speedrun.

Prime Agent is our open-source, self-improving coding harness built around two abstractions: the Recursive Language Model (RLM) and the Continual Harness. With Opus 5, it achieves 95.5% on ARC-AGI-3, surpassing the reported human expert baseline.

Introducing Prime Agent:A self-improving RLM harness for coding and long-running autonomous tasks.Designed to be both token-efficient and expressive through programmatic tool calling, context as a variable, multi-agent messaging, and a self-modifiable har

We ship first-party integrations for 23 agentic tasksets behind one taskset API - 365,000 software engineering, terminal, and search tasks, validated and ready for evals and RL training on Prime Intellect infrastructure.

We've raised $130M led by Radical Ventures, with participation from NVIDIA Ventures, Intel Capital, Dell Technologies Capital, and our existing investors - bringing our total funding to over $150M to build the open superintelligence stack.