Rapid progress on Epoch AI’s FrontierMath: Open Problems benchmark, with multiple AI-solved entries already marked since the first GPT-5.4 Pro success in March 2026, drives trader sentiment toward high implied probability of at least one more resolution by year-end. OpenAI’s GPT-6 Astra recently posted leading scores above 97% on Tier 4 v2 and solved two Erdős problems on the September-launched set, while Anthropic’s Claude Fable 5 achieved 88% on the hardest tier—outpacing prior GPT-5.6 variants. These results, alongside OpenAI’s broader internal claims of resolving over 100 open math problems, reflect accelerating large language model reasoning capabilities in verified, research-level mathematics. Traders monitor Epoch AI’s next official updates and any new model releases for catalysts that could confirm additional solved problems before December 31.