Anthropic’s Claude lineup, including Opus 5.5, Fable 5.1, and Mythos 5, currently leads Humanity’s Last Exam leaderboards with top scores of 61–67% across tool-using and no-tools variants as of mid-October 2026. Iterative releases emphasizing advanced reasoning, tool integration, and adaptive inference have driven consistent gains on the 2,500-question benchmark of expert-level academic problems. OpenAI’s GPT-6 series and Google’s Gemini models trail but remain competitive on certain subsets, while Meta and open-weight efforts lag further behind. With roughly two months left in 2026, additional Claude updates or refined evaluation harnesses could lift the highest score, though benchmark saturation and variant-specific rules introduce uncertainty for year-end resolution.