OpenAI's AI Solves the Navier-Stokes Problem

OpenAI said Tuesday its internal AI model produced a formally verified solution to the Navier-Stokes problem, one of math’s seven $1 million Millennium Prize Problems, after roughly 88 hours of training beginning August 28.

Key Takeaways

OpenAI said its AI model produced a formally verified solution to the Navier-Stokes problem after roughly 88 hours of training
Training began August 28, and the system was still actively running as of the announcement
The Clay Mathematics Institute designated the Navier-Stokes problem a Millennium Prize Problem worth $1 million in 2000
A formally verified proof is checked line-by-line by a computer proof assistant such as Lean or Coq

The claim, if validated by independent mathematicians, would mark the first time any AI system resolved a Millennium Prize question. The company has not yet published the full proof publicly, but stated the solution was generated by a system still actively running as of the announcement.

Details on the model architecture and verification methodology are expected in a forthcoming technical report from the research team.

Why The Navier-Stokes Problem Has Resisted Proof

The Navier-Stokes equations describe how fluids such as water and air move under the forces of pressure and viscosity. Formulated in the nineteenth century, they underpin virtually every simulation of fluid dynamics used in aerospace engineering, climate modeling, and cardiovascular medicine.

The core unsolved question is whether, given any smooth starting conditions in three dimensions, the equations always produce smooth, well-behaved solutions indefinitely, or whether they can “blow up,” developing infinite values in finite time.

Also Read: OpenAI Partner Status Officially Granted to Appinventiv

No mathematician has proven either outcome, which is why the Clay Mathematics Institute designated it a Millennium Prize Problem worth $1 million in 2000.

Formal verification adds a layer of rigor beyond a conventional mathematical proof. Rather than relying on peer reviewers to catch errors, a formally verified proof is checked line-by-line by a computer proof assistant, software such as Lean or Coq, that confirms every logical step conforms to foundational axioms.

This means machine-checkable certainty replaces human judgment, making it far harder for subtle errors to survive scrutiny.

If the result withstands that process, it would represent a landmark for both mathematics and AI-assisted scientific discovery.