OpenAI’s October 6 release of 372 result families—spanning algebra, number theory, topology, and theoretical computer science—from an unreleased internal frontier model has sharply elevated market expectations for AI-driven math advances in 2026. The dump follows September’s claimed progress on the Navier-Stokes Millennium Problem and includes Lean-verified proofs plus partial headway toward the Riemann hypothesis and Kakeya conjecture. Meta AI added five open-problem solutions earlier in the month via its consumer Muse Spark model. Traders now weigh how quickly other labs can match OpenAI’s scale, the pace of remaining Lean formalizations, and any year-end model releases or benchmark updates on suites like FrontierMath that could confirm or qualify additional breakthroughs before December 31.