**OpenAI’s October 6, 2026 release of an internal frontier model’s verified solutions to over 100 long-standing open math problems—including hundreds of Lean-checked manuscripts—has sharply raised expectations that Epoch AI will soon classify at least one additional FrontierMath: Open Problems entry as solved.** This follows earlier AI-driven resolutions (such as the Hadamard matrix problem) and builds on GPT-6 Astra’s dominance of FrontierMath leaderboards, where it reaches 97.6% on Tier 4 (v2) and has solved multiple Erdős problems that other models, including Claude variants, left untouched. Trader sentiment reflects OpenAI’s sustained lead in scaling reasoning compute and verification pipelines, though resolution hinges solely on Epoch’s official updates rather than third-party claims. Key near-term catalysts include further Epoch classifications and any additional lab announcements before the December 31 deadline.