OpenAI has announced a significant mathematical breakthrough, claiming its advanced AI models successfully resolved the 90-year-old Navier-Stokes problem in a mere 88 hours. This complex challenge, one of the Millennium Prize Problems, describes the motion of viscous fluid substances like water and air.
OpenAI's AI Effort to Solve Navier-Stokes
According to OpenAI, the endeavor began in late August after a new internal AI model demonstrated significantly enhanced mathematical capabilities. On September 1, 2026, spurred by rumors of other Millennium Prize Problems being solved, OpenAI deployed approximately 10,000 AI agents to tackle the Navier-Stokes equations.
These agents were organized into groups, each assigned different facets of the problem. Some groups focused on discovering a proof, while others searched for counterexamples. The system allowed agents to communicate and share insights, fostering a collaborative problem-solving environment.
By September 5, just 88 hours after the initial deployment, the AI system reached what OpenAI described as a resolution. During this intensive period, the agents exchanged an estimated 2.7 million messages and generated around 130 billion output tokens. The initial findings were then formalized and further verified over 17 hours using GPT-6 Astra, an advanced language model.
Controversy and Verification Concerns
Despite the audacious claim, OpenAI has explicitly stated it will not pursue the associated $1 million Millennium Prize. The company acknowledges that its work, while addressing the mathematical problem, requires rigorous independent scrutiny and has not yet been accepted by the Clay Mathematics Institute.
The announcement has also ignited a debate within the mathematical community. Tristan Buckmaster, a mathematics professor at NYU, along with Anthropic mathematician Levent Alpöge, had been working on similar problems using AI models, including OpenAI's Codex. Buckmaster raised questions about the speed at which OpenAI achieved its result after learning of their parallel research, suggesting potential influence from their earlier Codex sessions.
OpenAI has vehemently denied accessing or utilizing private user data for this project. The company asserted that no human or AI agent involved in the Navier-Stokes effort had access to user data, though it acknowledged that de-identified data can be used for general model improvements. OpenAI maintains that its proof was independently developed and fundamentally differs from the work of Buckmaster and Alpöge.
“We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics. The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.” – OpenAI (via X, September 8, 2026)
The resolution of the Navier-Stokes problem, if independently verified, would represent a monumental achievement in both mathematics and artificial intelligence, pushing the boundaries of what AI can accomplish in complex scientific domains.