Recent policy updates from Epoch AI on September 16, introducing a "human + AI" category for collaborative solutions, combined with the late-September marking of the Apéry-style irrationality proofs problem as solved, have reinforced trader expectations for additional FrontierMath: Open Problems resolutions. OpenAI's GPT-6 Astra leads v2 Tier 4 leaderboards at 97.6% while GPT-5.6 variants top Tiers 1-3, outpacing Anthropic's Claude Fable 5 series, though prior solves like the Ramsey hypergraph and Hadamard matrix involved multiple labs. With new Erdős problems added and verification relying on Lean formalization plus automated checkers, sentiment favors near-term progress ahead of September 30 and December 31 resolution windows, tempered by the benchmark's emphasis on verifiable, publishable advances over autonomous claims.