OpenAI Hints at Progress on a Second Millennium Prize Problem

OpenAI has revealed that its AI systems have made substantial progress on a second Millennium Prize problem, just days after the company announced an AI-generated result concerning the notoriously difficult Navier–Stokes problem. The company has not identified the second problem or said whether its work amounts to a complete solution. Instead, OpenAI said it is considering how to share the results carefully, creating anticipation within the mathematical and artificial intelligence communities.

The development follows OpenAI’s recent experiment in using large numbers of AI agents to work on major mathematical problems. According to OpenAI, its internal model coordinated groups of agents that could communicate, use tools and explore different approaches to the Millennium Prize problems. The company said thousands of concurrent agents were involved in the work that produced its Navier–Stokes result, demonstrating how increasingly capable AI systems can be organized to tackle problems that traditionally require highly specialized mathematical expertise.

OpenAI’s announcement is significant because the Millennium Prize problems were established by the Clay Mathematics Institute as a collection of seven exceptionally challenging mathematical questions, with a $1 million prize attached to each problem that is successfully solved according to the institute’s requirements. OpenAI’s recent Navier–Stokes work has not itself been described as claiming the Millennium Prize. The company has instead presented the result as evidence of progress in AI-assisted mathematical research and has emphasized the need for careful evaluation and verification.

The identity of the second problem remains unknown. Some online speculation has pointed toward the Hodge conjecture, another of the seven Millennium Prize problems, but OpenAI has not confirmed that this is the problem its systems are working on. The company’s wording is also important: saying it has made “substantial progress” does not necessarily mean that an entire problem has been solved. A mathematical result of this magnitude would require rigorous examination, verification and independent scrutiny before it could be treated as an accepted solution.

The announcement nevertheless highlights the changing relationship between artificial intelligence and advanced mathematics. OpenAI says its recent work represents a new stage in AI development, while acknowledging that major questions remain about how such systems should be evaluated, controlled and used responsibly. If the company eventually publishes its second result, mathematicians will likely examine not only the proposed mathematics but also the reasoning process, verification methods and relationship between AI-generated discoveries and human mathematical work. For now, OpenAI’s announcement leaves the identity and ultimate significance of the second problem unresolved.