Anthropic’s Claude Opus 5.5 and Fable 5.1 variants currently lead most standard HLE leaderboards at 61–67.7% as of mid-October 2026, reflecting strong performance on the 2,500-question expert benchmark spanning math, sciences, and humanities. OpenAI’s GPT-6 Astra posted an outlier 82.9% on the September-released HLE-Diamond subset with tools, highlighting how tool use and refined question sets can shift results. Rapid capability gains—roughly 30 points over the prior year—stem from iterative model releases, enhanced reasoning modes, and competitive pressure among frontier labs. Traders monitor official agi.safe.ai updates, new model cards from Anthropic, OpenAI, and Google, plus any year-end launches that could push verified scores past 70% before December 31 resolution.