OpenAI Says Agents Found a Lean-Checked Navier–Stokes Blowup—and Will Not Claim the Prize
On September 8, 2026, OpenAI published an AI-generated solution to Clay Navier–Stokes statements C and D: a finite-time singularity from smooth data, produced by ~10,000 concurrent agents and an unreleased model “significantly more capable than GPT-6 Astra,” then formalized in Lean. OpenAI will not claim the Millennium Prize.
TLDR
OpenAI on Tuesday, September 8, 2026 published “An OpenAI model proposes a solution to the Navier–Stokes problem.” Claim: an initially smooth fluid at rest, with a smooth force and finite energy, can form a finite-time singularity (Clay C and D). Internal model “significantly more capable than GPT‑6 Astra” (training from 28 Aug, still ongoing). 10,000 concurrent agents; solution 5 Sep ~88 hours after launch; 17 more hours of GPT‑6 Astra Lean formalization. NS effort: 2.7 million messages, ~130 billion output tokens; all problems: 4.9 million messages, ~300 billion tokens. Concurrent: Levent Alpöge (Anthropic) + Tristan Buckmaster (NYU) had forced Euler; OpenAI proved unforced Euler (100 agents, ~50 hours). OpenAI will not claim the Millennium Prize.
Caveats
This is a lab research publication, not Clay’s award. Independent mathematicians must check the Lean. Not a dual-file of Anthropic FLT (Sep 4)—different problem, different lab.
Product-line de-dupe: not Astra product (Sep 3), not Aug 1 ten math advances.
Why this story matters
Two labs in four days claimed machine-checked solutions to named millennium-class problems. Watch: referee reaction, and whether the unreleased post-Astra model becomes the next Critical card.
Sources
- OpenAI: An OpenAI model proposes a solution to the Navier–Stokes problem (September 8, 2026)
- Times of AI:
anthropic-claude-fermats-last-theorem-lean(September 4)