OpenAI says AI agents solved centuries-old mathematical problem in 88 hours

OpenAI has announced that it has solved a mathematical problem that has challenged scientists for nearly 200 years, after reportedly spending millions of dollars on the effort. The announcement, however, quickly sparked controversy after mathematician Tristan Buckmaster, a professor at New York University, accused OpenAI of accelerating its work on the problem after learning that he and another researcher from rival company Anthropic were close to announcing their own breakthrough.

Buckmaster also raised concerns about their ongoing research being stored in OpenAI’s Codex model, which is designed to write computer programs, potentially making the work visible to OpenAI’s team.

Buckmaster later clarified that he did not know what OpenAI’s model had done or how it had done it, and that he did not know whether their data had been used. He emphasized that he was not accusing anyone of wrongdoing.

Sebastian Bubeck, an OpenAI researcher, denied at a press conference on Tuesday that the company had used the researchers’ work or accessed materials shared with OpenAI servers. However, he acknowledged that the company could not rule out the possibility that data generated through the researchers’ use of its products had “helped improve our models.”

Solved in 88 hours

OpenAI said it tackled the Navier-Stokes problem using an internal system more powerful than its latest GPT-6 Astra model. According to the company, around 10,000 AI agents — systems capable of performing tasks autonomously — worked on the problem simultaneously and reached a solution within 88 hours.

“Our primary goal is to enable scientists to advance research and technology that benefits all of humanity,” the researchers wrote in a blog post. “We believe it is important to inform the world about the pace of AI progress and what can be expected from upcoming models.”

What is the Navier-Stokes problem?

The Navier-Stokes problem involves a set of differential equations developed by French scientist Claude-Louis Navier and Irish mathematician George Gabriel Stokes in the 19th century, with the equations dating back to 1822. They are used to describe the motion of fluids such as water, air and oil.

The equations play a critical role in a wide range of fields, from weather forecasting and the design of aircraft and ships to the study of ocean currents and blood flow through blood vessels.

The challenge is that scientists do not know whether the Navier-Stokes equations remain valid under all conditions or whether they can break down and produce infinite velocities. This is at the heart of the Navier-Stokes problem, which seeks to determine whether smooth, lasting solutions to the equations always exist.

OpenAI’s solution, however, suggests that the equations may actually break down under certain conditions and produce infinite velocities, a phenomenon known as a “blow-up.” According to the company, its GPT-6 Astra model took about 17 hours to verify the result.

Millennium Prize and AI safety concerns

Mathematicians who solve any of the seven Millennium Prize Problems are eligible for a $1 million award from the Clay Mathematics Institute. OpenAI, which is reportedly preparing for a potential public offering valued at up to $1 trillion, said it does not intend to claim the prize.

The reported breakthrough comes as OpenAI seeks to improve its public image following a period of growing concerns about the safety of its technologies. These concerns intensified after a security incident involving an external platform during testing last July.

The development also comes amid calls from some US senators for a permanent ban on the development of “superintelligent” AI systems capable of outperforming humans across all cognitive tasks.