OpenAI says it has used a new, unreleased internal AI model together with around 10,000 AI agents working in parallel to make progress on the Navier–Stokes existence and smoothness problem, one of six unsolved Millennium Prize problems concerning fluid dynamics that has stood for 90 years. The company describes the result as a milestone showing rapid improvement in AI capability, though it stresses the work has not been independently verified or accepted by the Clay Mathematics Institute, which administers the $1m prize, and it is not claiming the prize.

The AI agents took roughly 88 hours, exchanging almost three million messages and generating 130 billion output tokens, at an estimated cost of about $10m (£7.3m) based on OpenAI's own pricing. The effort resolved two of the four statements required for a full proof, and OpenAI says it began the work after hearing rumours that two Millennium Prize problems had been solved elsewhere. The claim has proved controversial: New York University mathematician Tristan Buckmaster alleges that he and an Anthropic researcher were independently working on the same problem using OpenAI's Codex tool, and that details of their progress reached OpenAI shortly before it began its own effort and published its results, a timeline OpenAI has not directly addressed in the article.