Breakthrough or dirty game? OpenAI claims AI solved a 90-year-old problem

OpenAI claims a new internal AI model, more powerful than GPT-6 Astra, has solved the Navier-Stokes problem, one of the Clay Institute's seven Millennium Problems. The company reports using 10,000 AI agents working in parallel for 88 hours. Meanwhile, a controversy has emerged: rival researchers claim the company tried to preempt their publication while attempting to exclude them from credit. OpenAI denies improper use of information.

OpenAI claims that a new internal artificial intelligence model, significantly more powerful than GPT-6 Astra, has succeeded in solving the existence and smoothness problem of the Navier-Stokes equations, one of the Clay Institute's seven Millennium Problems. According to the company, approximately 10,000 AI agents working in parallel exchanged 2.7 million messages and generated 130 billion tokens of output over 88 hours until the proof was obtained. The proof describes a fluid starting from rest, in which a vortex forms that contracts in width and elongates, and the flow velocity grows without bound in finite time. However, the proof uses a partial external force, a version that is not the intuitive core of the problem in the eyes of many mathematicians. Meanwhile, a controversy erupted: researchers Tristan Buckmaster and Levent Alpoge claim they had been working on a similar direction for months, and that OpenAI tried to preempt their publication while attempting to exclude Alpoge from credit due to his work at competitor Anthropic. Buckmaster claims he was offered two options for joint publication, and when he refused, was told "Why would you want to ruin your career?" OpenAI denies the allegations but admits it cannot completely rule out that information that had undergone de-identification from the two's use of its products was used to improve the models.