Anthropic's Claude Opus 5.5 and Fable 5.1 currently lead Humanity's Last Exam (HLE) leaderboards at 59–65% across tool-free and verified variants, outpacing OpenAI's GPT-6 Astra and GPT-5.5 series, which sit at 52–57% on independent evaluations as of mid-October 2026. Rapid progress on this 2,500-question expert benchmark—crowdsourced for frontier knowledge in math, science, and other domains—has lifted top scores roughly 30 points in a year, with tool-assisted runs showing even steeper gains for OpenAI models like Astra at 82.9% on HLE-Diamond subsets. Trader sentiment reflects Anthropic's sustained edge from recent releases alongside OpenAI's competitive but non-leading position, tempered by uncertainty over year-end model updates or evaluation protocol shifts that could alter the highest OpenAI score before December 31.