On Tuesday local time, OpenAI announced a major breakthrough that shocked the mathematical community, successfully providing a solution to one of the Millennium Prize Problems—the Navier-Stokes equations. These equations are extremely important in physics and engineering, mainly used to describe the motion of fluids such as air, water, and blood.

When disclosing technical details, OpenAI revealed for the first time that the key behind this breakthrough is an upcoming secret internal model that is significantly more powerful than GPT-6 Astra. Chart data shows that in a set of selected open math problems, the highest pass rate of this internal model reached nearly 48%, with performance three times that of Astra's best performance.

To solve this world-class problem, OpenAI invested millions of dollars in computing costs. In practice, the research team mobilized about 10,000 AI agents working in collaboration. After the first group of agents was launched, the entire solution process was efficiently completed within approximately 88 hours. OpenAI's research results indicate that three-dimensional fluid motion may produce singularities within a finite time, leading to the breakdown of the equations' description of fluids as continuous media.

However, this achievement also sparked controversy. Previously, mathematician Tristan Buckmaster from New York University and researcher Levent Alpöge from Anthropic had just published an AI-assisted study on related fluid dynamics equations. Subsequently, Buckmaster publicly expressed doubts, accusing OpenAI of rushing in after hearing about their research and adopting a rare research method that they had spent months developing.

In response to external criticism, OpenAI quickly clarified that its team had never accessed any related work through any means before the other party publicly released their findings, and absolutely did not access any specific external user data. At the same time, OpenAI emphasized that there were significant differences in the final proof methods between both sides.