Frontier Flux Original
On 8 September 2026 OpenAI published a research note saying a group of agents, using an unnamed internal model it describes as significantly more capable than GPT-6 Astra, produced an analytical proof and a Lean formalization that three-dimensional incompressible Navier-Stokes can develop a singularity in finite time. The company says it does not intend to claim the Clay Millennium Prize.

Abstract concept art for OpenAI Navier-Stokes research note
Frontier Flux / AI-generated
OpenAI published On the Navier-Stokes Millennium Prize Problem on 8 September 2026. It says an internal multi-agent system produced a proof that the three-dimensional incompressible Navier-Stokes equations can develop a singularity in finite time, plus a writeup PDF and a Lean formalization.
From the 8 September @OpenAI post: "We're sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics. The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra."
The research note says training of that unnamed internal model began 28 August 2026 and is still running. Agents launched 1 September after rumors that two Millennium problems had been resolved. A follow-up @OpenAI post says the internal model group arrived at the Navier-Stokes solution in 88 hours, using around 10,000 coordinating AI agents. The note adds Lean formalization and verification with GPT-6 Astra took another 17 hours, and that the Navier-Stokes group sent about 2.7 million messages and used about 130 billion output tokens.
OpenAI says the result is a smooth fluid at rest, with a smooth applied force and finite energy throughout, that develops a singularity in finite time. It frames this as statements C and D in the official Clay formulation: a vortex that spirals inward and elongates. It also reports an earlier unforced Euler regularity disproof (nearly 100 agents, about 50 hours), with a writeup PDF.
Key facts - OpenAI attributes the proof to an internal next-generation model, not to GPT-6 Astra as the solver. - OpenAI says it does not intend to claim the Clay Millennium Prize. - Concurrent work: OpenAI congratulates Levent Alpoge and Tristan Buckmaster, says they resolved forced Euler rather than Navier-Stokes, and that its researchers and agents did not see that work until it was public. It cannot rule out de-identified product-usage data helping improve models.
How to read this
These are OpenAI claims plus artifacts it posted. Clay has not awarded the prize. Independent mathematicians still have to check the writeup and the Lean. The applied force is part of the stated result.
Independence
Frontier Flux is independent. Facts track the 8 September @OpenAI posts and the official research note. Vendor Lean and token counts are OpenAI-reported.
Free
Free daily email of model releases, benchmarks, and lab notes from the last 24 hours across OpenAI, Anthropic, DeepMind, Meta, NVIDIA, and more. Capability-first, not company drama.
Opens Buttondown to finish signup. Confirm by email if asked. No ads. Unsubscribe anytime.