This paper reports on a breakthrough achievement by OpenAI: the resolution of the Navier–Stokes existence and smoothness problem, one of the seven Millennium Prize Problems. The research demonstrates that the dynamics of three-dimensional incompressible fluid motion, governed by the Navier–Stokes equations, can develop a singularity—a state where fluid velocity grows without bound within a finite time, even when the initial motion is smooth and the energy remains finite.

Core Problem

The Navier–Stokes equations describe the motion of viscous fluid substances:
$$\frac{\partial \mathbf{u}}{\partial t} + (\mathbf{u} \cdot \nabla)\mathbf{u} = -\frac{1}{\rho}\nabla p + \nu \Delta \mathbf{u} + \mathbf{f}$$
where $\mathbf{u}$ is the fluid velocity, $p$ is pressure, $\rho$ is density, $\nu$ is kinematic viscosity, and $\mathbf{f}$ is an external force. The fundamental question addressed is whether smooth, global-in-time solutions exist for these equations in three dimensions, or if they can develop a singularity. By establishing the existence of a singularity, the study confirms statements "C" and "D" of the Millennium Prize criteria, showing that the continuum approximation of fluid mechanics can break down under specific conditions.

Methodology

The solution was achieved through a multi-agent AI framework powered by an internal model superior to GPT-6 Astra. The process involved several technical stages:

• Multi-Agent Coordination: Approximately 10,000 concurrent agents were organized into groups tasked with exploring various formulations of the problem, including those implying proof of existence and those implying proof of blow-up (singularities).
• Iterative Refinement and Cross-Pollination: Agents were provided with tools including internet access and code execution capabilities. Insights were consolidated using Codex to merge useful intermediate results from different agent groups.
• Sequential Problem Solving: The agents first solved the unforced Euler equations (the limit of Navier–Stokes as viscosity $\nu \to 0$) by demonstrating a blow-up. This successful resolution provided a structural template for approaching the more complex Navier–Stokes case.
• Mathematical Proof and Verification: The agents derived an analytical proof identifying a specific vortex structure—a spinning fluid swirl—that undergoes inward spiraling and axial stretching. The researchers ensured the physical validity of the result by confirming that the total energy remains finite despite the localized velocity blow-up.
• Formalization: The analytical proof was formalized and verified using the Lean theorem prover, requiring 17 hours of additional processing time by GPT-6 Astra.

Conclusion

The result represents a significant milestone in AI capabilities, demonstrating that an autonomous multi-agent system can navigate deep mathematical frontiers. The researchers note that this discovery occurred concurrently with work by Levent Alpöge and Tristan Buckmaster on the forced Euler problem, though the methodologies and specific results between the two efforts remain distinct. OpenAI states that they do not intend to claim the Millennium Prize, presenting this work instead as evidence of the accelerating pace of AI development and as a benchmark for future research in scientific discovery.