Recent leaderboard updates show Anthropic’s Claude Opus 5.5 and Fable 5.1 reaching 60–65% on standard Humanity’s Last Exam variants, with Moonshot AI’s Kimi K3.1 Pro hitting 70% on tool-augmented versions as of mid-to-late October 2026. These gains reflect rapid progress in frontier large language models through improved reasoning chains, scale, and agentic capabilities, outpacing OpenAI GPT-6 variants and Google Gemini releases on most independent evaluations. The benchmark’s 2,500 expert-level questions across disciplines continue to leave headroom below saturation, so new model launches or refined tool use before year-end could push the highest score higher. Traders should watch Anthropic and OpenAI release timelines, as leadership shifts on verified runs often move implied probabilities quickly.