OpenAI’s early October 2026 release of 722 papers detailing solutions to 377 long-standing math problems, many formalized in Lean, has sharply accelerated progress on research-level benchmarks. Building on its September claim of solving the Navier-Stokes equations, the unreleased internal model produced results across algebra, number theory, and topology, including partial advances on Millennium Prize problems such as the Riemann hypothesis. Epoch AI’s FrontierMath Erdős set and Tier 4 tasks have already seen multiple verified solutions by models like GPT-6 Astra since August, with traders viewing the latest wave as likely to trigger additional official “solved” designations. Near-term resolution hinges on Epoch’s verification pace amid rapid model iteration and competing efforts like doubleAI’s graph-theory advance.