OpenAI’s early October 2026 release of 722 manuscripts solving or advancing 377 previously unsolved math problems—including claimed progress toward Millennium Prize challenges like Navier-Stokes and partial results on the Riemann hypothesis—has sharply lifted trader expectations for additional FrontierMath open-problem resolutions. The company’s unreleased internal model generated verified Lean proofs at scale, averaging roughly three hours of compute per result, outpacing prior benchmarks where GPT-6 Astra already leads Epoch AI’s Tier 4 leaderboard at 97.6%. This surge follows earlier AI-solved entries on the FrontierMath list and intensifies competition with Anthropic’s Claude models, though independent verification and Epoch AI’s formal classification remain the market’s sole resolution trigger. Near-term catalysts include further model attempts, proof audits, and any new Epoch announcements on remaining high-notability problems.