OpenAI’s early October 2026 release of 722 papers detailing solutions to 377 previously unsolved math problems—many verified in Lean and spanning number theory, algebra, and topology—has sharply elevated expectations for rapid AI progress on frontier-level challenges. This follows the company’s September claim that an internal model solved the Navier-Stokes Millennium Prize problem. On Epoch AI’s FrontierMath benchmark, OpenAI’s GPT-6 Astra leads with near-saturation scores on the hardest Tier 4 problems, outpacing competitors like Claude variants. Traders are watching for independent verification of the new results, potential contributions to remaining Millennium problems such as the Riemann hypothesis, and responses from other labs or startups like doubleAI that are also reporting breakthroughs on long-open questions.