It took OpenAI's agents 130 billion tokens to crack a 90-year-old math problem

OpenAI's AI agents have reportedly solved a 90-year-old math problem, the Navier-Stokes equations, in a feat that cost over $10 million, according to estimates. The company's agents sent 2.7 million messages and used 130 billion output tokens to crack the problem, which had been offered a $1 million prize by the Clay Mathematics Institute.

While OpenAI CEO Sam Altman joked that AI was a bubble, the news has been celebrated within the AI industry, though it has also been met with criticism. Two mathematicians, Tristan Buckmaster and Levent Alpöge, questioned whether OpenAI used their research data to train its models, though the company denied direct access and only stated that de-identified data may have helped improve its models.

The verification process took an additional 17 hours to confirm the solution, which was found after 88 hours of intensive AI use involving 10,000 concurrent agents.