OpenAI’s October 6–7, 2026 release of 722 AI-generated math papers from an unreleased internal frontier model, including verified progress on open problems in number theory, algebraic geometry, and partial advances toward Millennium Prize questions such as the Riemann hypothesis, has sharply lifted trader expectations for additional FrontierMath Open Problems being marked solved. GPT-5.6 Sol and GPT-6 Astra variants currently lead Epoch AI’s corrected FrontierMath v2 leaderboards at 89% and 97.6% on Tiers 1–3 and Tier 4 respectively, reflecting superior iterative reasoning and Lean formalization capabilities. Mathematicians are still verifying the new results, while Epoch AI’s independent classification process and potential new problem additions remain the direct resolution triggers; other labs trail significantly on the same benchmarks.