OpenAI announced that an internal artificial intelligence system solved the Navier–Stokes existence and smoothness problem, one of the seven Millennium Prize Problems. According to the company’s statement, roughly 10,000 agents were used, the computation ran for 88 hours, and about 300 billion tokens were consumed. The result was produced as a formal proof in Lean.
The announcement also states that the model behind this run exceeds the capabilities of GPT-6 Astra. OpenAI framed the claim as part of a series of recent achievements, referencing ARC-AGI-3 at 99.9%, ExploitBench at 100%, and an Erdős conjecture reportedly solved earlier in the year.
Practical implications
OpenAI’s release suggests a change in tempo for addressing long-standing mathematical problems: benchmarks are improved, papers are produced, public announcements follow, and new targets enter training pipelines. The company appears to be treating Millennium Prize Problems as a prioritized backlog; there are seven such problems in total. The implication is that tasks which took decades or centuries for the mathematical community to erect as challenges may now be tackled in much shorter timescales by automated agent systems.
What the announcement does and does not say
The company provided specific operational figures (10,000 agents, 88 hours, 300 billion tokens) and described the outcome as a Lean formal proof. It also claimed the underlying model outperforms GPT-6 Astra. The announcement does not, however, include details in this text about independent peer review, external verification of the proof, or release of the full technical paper and datasets for replication.
Why this matters
The Navier–Stokes existence and smoothness problem is a central open question in mathematical analysis and fluid dynamics. If the Lean formal proof is accepted by the mathematical and formal-methods communities, it would represent a major milestone for both mathematics and automated theorem proving. Until independent validation and detailed publication are available, the broader community will need to assess the claim through standard scholarly scrutiny.
Closing note
OpenAI’s communication follows a pattern of rapid claims about solving historically difficult problems. The next steps that will determine the claim’s standing are the release of detailed technical materials and independent expert verification.
