OpenAI's Secret AI Solves $1M Math Problem

OpenAI says an unreleased internal model just cracked one of the seven Millennium Prize problems, a set of math puzzles that have sat unsolved for decades with $1 million attached to each. The target this time was Navier-Stokes, the equations that describe how fluids move, and the kind of problem that's chewed up careers without producing a proof. According to OpenAI, the work involved roughly 10,000 AI agents running at once for 88 hours, on a system the company describes as "significantly more capable" than GPT-6 Astra, which shipped less than a week ago. Sam Altman put the compute bill in the millions of dollars and called the result one of the most amazing moments of his career.

Not everyone is convinced the credit belongs entirely to the machine. NYU mathematician Tristan Buckmaster, who has spent roughly a year chasing the same proof, is asking whether OpenAI's model had access to his own drafts through Codex, OpenAI's coding tool. That question, and how OpenAI answers it, is where the story gets interesting.