Recent releases of OpenAI’s GPT-6 Astra in early September and Anthropic’s Claude Fable 5.1 have driven sharp gains on math-focused leaderboards, with expected performance scores now reaching the low-to-mid 80s percent range across rigorous, contamination-resistant benchmarks like ArXivMath and proof-based competitions. These frontier large language models demonstrate stronger multi-step reasoning and theorem-proving than prior generations, narrowing the gap to human expert levels on several tracks. Competitive pressure between OpenAI and Anthropic, alongside incremental updates from Google and open-weight efforts, keeps momentum high heading into the final months of 2026. Traders monitoring release cadence and benchmark thresholds see these advances as the dominant factor supporting elevated implied probabilities for further score gains by year-end.