OpenAI’s AI Agents Take On a Legendary Fluid Mathematics Problem

OpenAI says an AI system has found a solution to the Navier-Stokes problem, one of the most difficult challenges in modern mathematics. The announcement, made on September 8, 2026, describes a system that used roughly 10,000 AI agents, 130B tokens, and 88 hours of work.
The claim has already raised questions about how the result should be evaluated and who deserves credit. OpenAI describes the work as a collaboration among thousands of agents, rather than a discovery made by one model working alone. The result also needs a rigorous review before anyone can treat it as a confirmed solution to the Millennium Prize Problem.
How the AI system approached the problem
The Navier-Stokes equations describe how fluids move in space. They are used to study the behavior of liquids and gases, but the mathematics behind them becomes difficult when researchers try to prove whether certain solutions remain smooth or develop singularities called blow-ups.
OpenAI first set 1000 AI agents to work on the Euler problem, a related mathematical problem. Those agents took 50 hours to find blow-ups. OpenAI then set 10,000 agents to extend the blow-up result to the full Navier-Stokes problem, a stage that took 11 hours.
OpenAI announced that blow-ups can appear in Navier-Stokes equations, similar to Euler equations. That finding addresses the central issue connected to the Navier-Stokes Millennium Prize Problem, for which a correct solution comes with a $1 million reward.
The full effort took 88 hours, according to OpenAI’s announcement. Its system used parallel test-time compute, sending many lines of mathematical exploration forward at once, while also allowing the agents to organize their work among themselves.
“The Navier Stokes solution was the result of a collaboration of ~10,000 agents working together,” Ethan Knight said.
The agents were trained over roughly a year with multi-agent reinforcement learning. OpenAI said the AI model behind the work was “significantly more capable” than GPT-6 Astra, and the system used Astra-next for the reported effort.
Why the claim matters — and why review comes next
OpenAI’s announcement places the result in a long history of attempts to understand the Navier-Stokes equations. Venkat Chandrasekaran described the difficulty in direct terms: “This problem has remained unsolved for 200 years because the Navier-Stokes equations are just so enormously complex, and the pen-and-paper calculations you need to do in order to solve this problem are just mind-bogglingly intricate.”
That complexity is also why the announcement alone does not settle the question. A proposed solution must show that every step works, address the exact conditions of the problem, and survive close examination by mathematicians. The process involves more than producing an answer that looks convincing or passes an internal test.
Martin Bridson emphasized that point: “The process of evaluation is deliberately unhurried, and we shall ensure that it is absolutely rigorous.”
The result has also created controversy over the process and attribution. If thousands of agents explored the problem together, the usual idea of a single author becomes harder to apply. The work involved OpenAI researchers and an AI system built from coordinated agents, which raises a practical question: should the discovery belong to the organization, the researchers, the model, or the entire system that produced the proof?
The scale of the effort adds another layer. OpenAI said it would cost around $15 million to run the same problem. The project is also associated with 130B tokens and a $40M figure, showing how much computing and model activity went into the attempt.
A new step in AI-led mathematics
OpenAI’s Navier-Stokes claim follows other AI-led mathematical discoveries, including cracking a conjecture by Paul Erdős and formalizing Fermat’s last theorem. Those examples helped show how AI systems can assist with problems that demand long chains of reasoning, formal checks, and repeated exploration.
The Navier-Stokes work pushes that approach toward a problem involving continuous motion, fluid behavior, and singularities. Instead of asking one model to produce a final response, OpenAI used many agents that could test ideas in parallel and build on one another’s work.
That model of cooperation may be the most important part of the announcement, even before the mathematics receives final approval. The system was not described as a single artificial mathematician. It was a network of agents trained to work together, compare paths, and extend partial results.
Still, the central claim remains a claim until the evaluation is complete. If the proof survives rigorous review, OpenAI’s work could become a contender for the second awarded solution among the Millennium Prize Problems. If parts fail, the process may still show where AI systems can contribute to difficult mathematical research.
For now, the announcement offers both a remarkable result and a challenge to traditional ideas about mathematical discovery. OpenAI says its agents found the Navier-Stokes blow-up result in 88 hours. The next test belongs to human evaluation.
Based on
- [AINews] OpenAI reports Navier-Stokes singularity find in 88 hours using Astra-next, roughly 10,000 agents and 130B tokens (>$40M), a contender for second ever Millennium Prize awarded — latent.space
- OpenAI has solved the Navier-Stokes Millennium problem using $15m of AI effort | New Scientist — newscientist.com
- OpenAI Cracks 200-Year-Old ‘Millennium’ Math Problem—But Credit Dispute Looms — forbes.com




