OpenAI Claims AI Solved Navier-Stokes in Days, Sparking Credit Dispute

OpenAI announces breakthrough on century-old math problem

OpenAI announced Tuesday that one of its unreleased AI models solved the Navier-Stokes problem, a mathematical puzzle that has stumped experts for generations. The company said the achievement took approximately 88 hours and required millions of dollars in computing resources.

The Navier-Stokes problem is one of seven Millennium Prize Problems identified by the Clay Mathematics Institute as the hardest open questions in mathematics. Each carries a $1 million reward. The equations at the heart of the problem describe fluid motion—how air and water flow—and have applications in aircraft design, weather forecasting, and blood flow analysis. OpenAI's work purportedly demonstrates that these equations can break down over time, answering the specific question posed by the prize.

Why it matters

If validated, this represents a fundamental shift in how humanity tackles problems once considered solvable only through human mathematical intuition. The computational cost—roughly 1,000 times what OpenAI spent on previous mathematical work—signals that solving the world's hardest problems may require industrial-scale AI infrastructure. For business leaders, this underscores the strategic importance of compute capacity and raises questions about the return on investment for frontier AI research.

Massive computational effort deployed

According to OpenAI, the company began training a new, more capable model in late August. When online rumors suggested a competitor had solved Millennium problems, OpenAI directed the system toward all six remaining unsolved challenges. The Navier-Stokes problem showed unexpected promise, prompting the company to concentrate its resources there.

In the final stage, 10,000 AI agents—autonomous programs operating independently—worked simultaneously on the problem, exchanging information among themselves. OpenAI researcher Sebastien Bubeck confirmed the computing costs ran "emphatically in the millions of dollars." Mark Chen, OpenAI's chief research officer, characterized the achievement as "a significant milestone for AI research" that demonstrates "even more of our hardest questions would become possible to answer."

OpenAI stated it will not claim the monetary prize if its solution is confirmed.

Controversy over credit and approach

The announcement arrived hours after Tristan Buckmaster, a mathematician at New York University, and Levent Alpoge, who works at OpenAI competitor Anthropic, released their own AI-assisted work on three related equations. Buckmaster issued a statement alleging that OpenAI only pursued the problem after word of his research spread and adopted the same unusual approach he and Alpoge had developed over months.

OpenAI researchers denied accessing the rival team's work and said they offered to share credit before determining that Buckmaster and Alpoge had not fully solved the problem. "To be clear, we did not use their prompts or proofs to prompt our models or direct our agents," Bubeck said. However, OpenAI later acknowledged on X that "while unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models."

Validation process ahead

Under the Millennium Prize rules, a proposed solution must be published in a peer-reviewed journal and survive two years of acceptance within the mathematical community before the Clay Mathematics Institute will convene a committee to evaluate it. Professor Martin Bridson, President of the Clay Mathematics Institute, told reporters that "the process of evaluation is deliberately unhurried, and we shall ensure that it is absolutely rigorous."

Only one Millennium Prize Problem has been solved since the list was published in 2000.

These details were first reported by Alex Pigman for Yahoo News.