OpenAI’s GPT-6 Astra recently posted 97.6% on FrontierMath Tier 4 (v2) and solved two Erdős problems in early benchmark runs, reinforcing its lead after saturating the corrected private set. The September 2026 claim that internal agents resolved key aspects of the Navier-Stokes existence and smoothness problem in 88 hours using roughly 10,000 agents further elevated trader expectations for OpenAI on future “another problem solved” resolutions. Anthropic’s Claude Fable models trail on public leaderboards but have independently solved select open problems, while Google’s Gemini entries remain lower. Key near-term catalysts include ongoing FrontierMath Erdős evaluations and any verified solutions to additional open research questions, where model access, test-time compute, and verification tooling determine outcomes.