OpenAI Claims 10,000 AI Agents Cracked a $1 Million Millennium Math Problem in 88 Hours

The Navier-Stokes existence-and-smoothness problem, unsolved for nearly a century, may have fallen to an army of autonomous agents — but a priority dispute with rival mathematicians is already brewing.

For a few hours on Tuesday, September 8, the quiet frontier of pure mathematics collided head-on with the AI race. OpenAI announced that a swarm of roughly 10,000 autonomous agents — running on an internal model the company says is "significantly more capable than GPT-6 Astra" — had cracked the Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize Problems the Clay Mathematics Institute staked with $1 million bounties back in 2000.

The equations, first written down in the nineteenth century by Claude-Louis Navier and George Gabriel Stokes, describe how fluids move and underpin everything from aircraft design to weather forecasting and the study of blood flow. For roughly 90 years, mathematicians have been unable to say whether a smooth, incompressible fluid could spontaneously "blow up" into a singularity — a point where its speed climbs without bound in a finite time.

OpenAI's agents claim to have found that singularity. After about 88 hours of work, the group produced a proof of a vortex that spirals inward, stretches like spaghetti, and accelerates while keeping its energy finite. A second model then spent 17 more hours formalizing the result in Lean, a programming language that mechanically verifies every step, Quanta Magazine reports.

But the breakthrough arrives tangled in a priority dispute. Hours earlier, NYU mathematician Tristan Buckmaster and Anthropic's Levent Alpöge announced closely related results on a sibling equation, and both camps built on the techniques of Diego Córdoba and Luis Martínez-Zoroa. OpenAI says it recognized the pair's priority on the "forced Euler" problem and reached out for a concurrent release, Interesting Engineering notes — even as Buckmaster suggested OpenAI's agents may have benefited from work done on its own models. For its part, OpenAI says it does not intend to claim the prize itself.

If the result survives peer scrutiny, it would be the most significant mathematical proof ever reached by an AI system — and a clear preview of what lies beyond GPT-6 Astra.