While OpenAI acknowledged that de-identified data from NYU professor Tristan Buckmaster and Anthropic employee Levent Alpöge may have been involved in training the models, the company stated that its agents did not access the duo's specific work before completing the proof.
This breakthrough signifies a major leap in AI infrastructure and reasoning capabilities, as the solution was generated by a coordinating system of approximately 10,000 concurrent agents.
These agents utilized 130 billion output tokens—units of text processed by the model—to navigate complex fluid dynamics, which are essential for aircraft design and weather forecasting.
OpenAI reported that the internal model used is significantly more powerful than its GPT-6 Astra product, suggesting that the pace of capital expenditure in AI development is yielding systems capable of solving high-level scientific problems that have eluded human researchers for decades.
The resolution involved both an analytical proof and a formalization in Lean, a specialized programming language used for mathematical verification.
In addition to the Navier–Stokes result, the system resolved the unforced Euler regularity problem, a related question regarding fluids without viscosity or external force.
Although OpenAI does not intend to claim the Millennium Prize, the company indicated that this milestone marks a new era of steerable AI research where multi-agent systems can be used to accelerate technological discovery and scientific inquiry.