On Tuesday, OpenAI published a proof that is intended to solve the so-called Navier-Stokes problem. This is one of the seven Millennium Problems in mathematics, for the solution of which the Clay Mathematics Institute has offered a prize of one million US dollars each. The official problem description dates back to the year 2000.
According to OpenAI, around 10,000 AI agents worked in parallel on Navier-Stokes at times. After about 88 hours, they had found the proof, and another 17 hours were needed for its formalization and verification. For Navier-Stokes alone, the agents exchanged 2.7 million messages and generated around 130 billion output tokens. OpenAI estimated the computing power used for this to Wired at several million US dollars. The basis for the agents was an internal model, not yet published, which according to OpenAI is significantly more powerful than the new top model GPT-6 Astra.
An independent review of the 166-page paper is still pending. Since the regularity question of the Navier-Stokes equations has occupied mathematicians for around 90 years, a proof achieved with the help of AI would be a mathematical breakthrough and, at the same time, a significant success for the use of AI in research.
