OpenAI Solves Millennium Prize Math Problem Using 10k Agents and 130B Tokens

OpenAI deployed a system of 10k AI agents to successfully solve one of the Millennium Prize Problems. The massively parallel operation consumed 130B tokens over an 88-hour computing window. Scaling agents in parallel yields slightly sublinear speedups, meaning a 16-agent cluster solves a task faster but at a higher total token cost than a single agent. Researchers are hitting latency bottlenecks with sequential reasoning models, forcing a structural shift toward multi-agent clusters to buy faster test-time answers. "The reality is that OpenAI has trained a very powerful model. We can get that model to operate over very long horizons. We can get it to think in parallel." — Noam Brown