OpenAI announced today that an internal AI system has produced what it describes as a solution to the Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize Problems established by the Clay Mathematics Institute in 2000. Each problem carries a $1 million prize. Only one has ever been solved.
The company says the proof was generated by a model significantly more capable than GPT-6 Astra, its current flagship. It has not released the model publicly.
What the Proof Claims
The Navier-Stokes equations, dating to the 19th-century work of Claude-Louis Navier and George Gabriel Stokes, describe the motion of viscous fluids. For roughly 90 years, mathematicians have debated a fundamental question: can smooth three-dimensional fluid flow break down into singularities, or does it stay smooth forever?
OpenAI says its system demonstrates the first outcome. According to the company, an initially smooth fluid at rest, subjected to a smooth external force, can develop a singularity in finite time while its total energy remains bounded. The solution describes a vortex that spirals inward, speeding up in a controlled manner until it reaches a singular point.
The company released a 165-page writeup alongside a formalization in Lean, the proof assistant mathematicians use to verify each logical step mechanically. OpenAI states it does not intend to claim the prize money.
The Scale of the Effort
The proof emerged from an orchestrated swarm of AI agents. OpenAI says roughly 10,000 concurrent agents worked on the Navier-Stokes problem, reaching their result in about 88 hours after exchanging 2.7 million messages and consuming approximately 130 billion output tokens. Lean formalization and verification took an additional 17 hours, handled by GPT-6 Astra rather than the unreleased model.
Computing costs reportedly ran into the millions of dollars. The company said it began its push on September 1 after hearing a rumor of progress on Millennium Prize problems. A smaller group of about 100 agents also resolved the unforced Euler regularity problem as an intermediate step.
A Parallel Controversy
The announcement lands amid a dispute over credit. NYU mathematician Tristan Buckmaster and Anthropic researcher Levent Alpöge had been working for roughly a year on related fluid-dynamics questions. On the night before OpenAI's announcement, Buckmaster posted that he and Alpöge had proven blow-up for the Euler equations under smooth forcing, building on work by Diego Córdoba and Luis Martínez-Zoroa.
Buckmaster published a statement alleging that OpenAI learned of their progress and pivoted resources toward the full Navier-Stokes problem. He says he was presented with publication options that would have excluded Alpöge from authorship because of his Anthropic affiliation. OpenAI mathematician Sébastien Bubeck denied the allegations in a press briefing, stating that the company's internal model solved the Euler problem through entirely different methods.
OpenAI acknowledged it contacted Buckmaster and Alpöge after completing its own proof, at which point it learned their work addressed forced Euler equations rather than Navier-Stokes. The company said it recognizes the priority of their result on that distinct problem.
What Happens Next
Whether OpenAI's proof withstands scrutiny is now the question that matters. The mathematical community will examine the Lean formalization. Acceptance by the Clay Mathematics Institute would be required for any prize award. The only previous Millennium Prize solution, Grigori Perelman's proof of the Poincaré Conjecture in 2002-2003, took years to verify and the award itself to be offered.
The pace of discovery raises its own questions. Four days from rumor to proof is a research tempo that did not exist two years ago. If the result holds, it will be remembered as the first Millennium Prize problem solved primarily by machine. The implications for mathematical research, and for who gets credit when AI produces breakthroughs, remain unsettled.
