OpenAI Says AI Solved Navier–Stokes in 88 Hours

Overview

• OpenAI announced Tuesday that a nonpublic next‑generation model produced a claimed solution to the Navier–Stokes Millennium Prize problem after about 88 hours of computation using more than 10,000 cooperating agents, roughly 130 billion generated tokens, and what the company described as millions of dollars in compute costs.
• OpenAI posted a detailed account of the model’s reasoning and said it is not claiming the Clay Mathematics Institute’s million‑dollar prize for this result, inviting outside mathematicians to examine the work.
• Two researchers, Tristan Buckmaster and Levent Alpöge, accused OpenAI of reproducing ideas from their unpublished work; OpenAI first denied using their prompts or outputs and later said it could not fully rule out that anonymized data influenced model training.
• The Clay Mathematics Institute and leading mathematicians stressed that formal acceptance and prize consideration require lengthy peer review and publication, so the claim will stand only after extended independent verification.
• The episode highlights rapid AI progress in formal mathematics, raises new questions about data provenance and authorship norms, and intensifies competition among labs while prompting concern about how automated proofs will be verified and taught.