OpenAI says internal model solved Navier-Stokes problem

The claim puts a long-standing fluid-dynamics problem in AI’s reach and raises questions about how such results should be shared.

OpenAI says an internal model solved the Navier-Stokes problem, one of the seven Millennium Prize Problems, after training began on August 28 and the company used 10,000 concurrent agents to reach the result. That matters because the problem has been open for about 90 years and sits at the center of a major area of applied mathematics around fluid flow. The company also says the model is more capable than GPT-6 Astra, and that it plans to release more math results generated by the system. The announcement has triggered concern among mathematicians, who are now also questioning OpenAI’s handling of disclosure, coordination, and its relationship with the field. OpenAI has responded by setting up an independent advisory group of mathematicians to help guide how it presents and releases future results.