OpenAI Says AI Agents Solved $1M Navier-Stokes Math Problem In 88 Hours, Sparking Skepticism Among Experts

SAN FRANCISCO — OpenAI announced Tuesday that an internal, unreleased artificial intelligence system had produced a solution to the Navier-Stokes existence and smoothness problem, one of mathematics' seven Millennium Prize Problems, using roughly 10,000 AI agents working together over approximately 88 hours, a claim that has since drawn both fascination and pointed skepticism from the mathematics community.

The Navier-Stokes equations describe how fluids such as water and air move and are commonly used to predict weather patterns. The unsolved question at the center of the Millennium Prize challenge asks whether smooth, well-behaved three-dimensional fluid motion can eventually break down and produce speeds that grow without limit, a phenomenon mathematicians refer to as finite-time blowup. OpenAI's system concluded that it can, describing a specific configuration in which a fluid vortex tightens and spins ever faster while the fluid's total energy remains finite throughout the process.

In its official announcement, OpenAI described the achievement in terms of the broader trajectory of its technology.

"We're sharing a solution to the Navier-Stokes existence and smoothness problem, one of the Millennium Prize Problems," the company wrote. "This proof, produced by an internal OpenAI system, shows that the dynamics of the Navier-Stokes equations for fluid motion can develop a singularity in finite time."

The Clay Mathematics Institute named the Navier-Stokes problem as one of seven Millennium Prize Problems in 2000, with each carrying a $1 million award for a verified solution. Before Tuesday's announcement, only one of the seven problems, the Poincaré conjecture, had ever been solved. OpenAI has said it does not intend to claim the $1 million prize for its result.

According to reporting from Quanta Magazine, OpenAI's process began by deploying smaller groups of AI agents against related but somewhat easier problems. The company said nearly 100 agents worked together for approximately 50 hours to produce a disproof related to the Euler equations, a related but distinct mathematical framework, before OpenAI escalated the effort by deploying roughly 10,000 agents specifically against the Navier-Stokes problem. After 88 hours, those agents had produced a proof of a singularity in the Navier-Stokes equations, and an additional 17 hours of work by another AI model formalized that result using a verification system called Lean. In total, the agents involved in the effort exchanged nearly 5 million messages and generated approximately 300 billion output tokens across all attempted problems, with the Navier-Stokes-specific work alone involving 2.7 million messages and roughly 130 billion output tokens.

OpenAI researcher Sébastien Bubeck offered a sweeping characterization of the result's significance, according to The New York Times.

"This is a spectacular culmination of the arc we have seen over the past twelve months," Bubeck said.

The announcement has not gone without controversy. According to Quartz's reporting, OpenAI said its effort began Sept. 1 after the company heard rumors that two researchers, Levent Alpöge, an Anthropic employee, and Tristan Buckmaster, a mathematics professor at New York University, had already resolved a related problem. After completing its own proof and Lean verification on Sept. 6, OpenAI said it reached out to the two researchers to propose a joint announcement, only to learn that their work actually addressed a related but distinct problem involving the forced Euler equations rather than Navier-Stokes itself. OpenAI said it recognizes the priority of the two researchers' work on that separate problem, and maintained that its AI agents did not access the researchers' work through any means before it was made public, though the company acknowledged it could not entirely rule out indirect exposure to de-identified information related to the research.

Terence Tao, a prominent mathematics professor at the University of California, Los Angeles, offered a more measured perspective on the broader significance of AI-assisted breakthroughs of this kind, according to reporting cited by ETV Bharat.

"These questions are lighthouses," Tao said. "They are great focus points that attract the efforts of human scientists."

Tao separately cautioned that reliance on AI systems to solve landmark mathematical problems carries risks for how mathematicians themselves develop deep understanding of their field, noting that the sustained effort required to work through difficult problems often teaches researchers something valuable in the process, comparing the experience to the value of physical exercise pursued toward a specific fitness goal rather than simply outsourcing the effort entirely.

Despite the speed and scale of OpenAI's announced result, the broader mathematics community has not formally accepted the proof as valid. According to IBTimes UK, OpenAI's solution addresses two of the four formal statements required under the Clay Mathematics Institute's official problem formulation, and the proposed solution has not yet undergone independent verification or been accepted by the institute, which oversees the Millennium Prize Problems and their associated awards. Rigorous verification of a proof addressing a problem of this complexity and significance typically takes considerable time, often spanning at least two years, given the need for the broader mathematical community to carefully scrutinize every step of the argument before a result of this magnitude can be considered formally established.

OpenAI has framed the demonstration primarily as evidence of rapid progress in AI's capacity to tackle complex scientific and mathematical challenges, describing a core goal of its research as empowering scientists to advance work that benefits humanity broadly, rather than positioning the result as a final, independently verified mathematical breakthrough in its own right.

With the proposed solution now circulating publicly alongside both its formal writeup and Lean-based formalization, mathematicians across the field are expected to spend the coming months, and likely years, scrutinizing the underlying argument in detail. Whether OpenAI's claimed resolution of the Navier-Stokes existence and smoothness problem ultimately withstands that extended review process, and what it might reveal about the current capabilities and limitations of AI-driven mathematical research more broadly, remains an open question that the broader mathematics community has only just begun to address.