OpenAI’s October 2026 release of 722 AI-generated manuscripts from an unreleased internal frontier model, covering hundreds of open problems in number theory, topology, and theoretical computer science, has intensified trader focus on FrontierMath Open Problems. Many results include Lean formalizations, building on earlier claims such as progress toward the Navier-Stokes Millennium problem and multiple Erdős-type conjectures. Epoch AI’s official classification drives resolution, independent of third-party verification or company announcements, creating uncertainty around which solutions will receive benchmark credit. OpenAI’s GPT-6 series leads current FrontierMath leaderboards, ahead of Anthropic’s Claude Opus and DeepMind’s AlphaProof systems, while ongoing model iterations and potential new problem additions through year-end remain key catalysts that could shift the probability of additional official solves.