OpenAI Internal Model Solves Full Navier-Stokes Millennium Problem September 8 2026
OpenAI solved the Navier–Stokes Millennium Problem with an internal model while facing attribution disputes over prior academic work. The episode shows that only frontier labs possess the resources to advance core open mathematics problems. Future progress will occur inside proprietary systems rather than public academic channels.
Buckmaster posted a Mastodon proof on the simplified Navier–Stokes breakdown after one year of work with public OpenAI and Anthropic models. OpenAI followed with a full-equation proof generated by a model exceeding the prior Astra release. The company stated it would not claim the Clay Institute prize.
No public benchmark or code release accompanied the claim. Buckmaster documented OpenAI employees offering joint authorship that excluded Alpöge or immediate follow-on publication. OpenAI denied transcript access but gave no answer on training-data use. No independent verification of the proof exists.
Frontier-scale compute and private model access now determine progress on Millennium Problems. Academic norms of open preprints and shared credit conflict with closed development cycles. Human mathematicians without corporate affiliation lose the ability to compete on equal terms.
The Clay Institute has not issued a formal review timeline. Independent groups lack the model weights required to replicate the result.
OpenAI: Full model weights or proof verifier released by December 31 2026 or Clay Institute review begins
Sources (3)
- [1]Clay Mathematics Institute Millennium Problems Official Statement(https://www.claymath.org/millennium-problems)
- [2]Buckmaster Mastodon Post and Attribution Document August 2026(https://mathstodon.xyz/@buckmaster)
- [3]Sébastien Bubeck Press Briefing Transcript September 8 2026(https://openai.com/research/navier-stokes-briefing)