INTELLECT-2 Decentralized RL Training Recipe Resurfaces

Prime Intellect's INTELLECT-2, the first globally decentralized RL training effort that produced a 32B reasoning model, is back in discussion. Its recipe pools heterogeneous GPUs with asynchronous weight broadcasts and Toploc-based contribution verification.

2026-10-14 ~ 2026-10-14 · 2 related posts

• INTELLECT-2 distributed RL recipe from May 2025 resurfaces with full system architecture — willcb · 2026-10-14
• INTELLECT-2 Revisited: The Open Recipe Behind the First Globally Decentralized 32B RL Training Run — willcb · 2026-10-14