Anthropic’s Claude series dominates recent HLE leaderboards through iterative releases emphasizing reasoning, tool use, and agentic capabilities, with Claude Opus 5.5 posting the highest verified scores around 61% on independent evaluations and up to 67.7% in lab-reported tool-augmented runs as of October 2026. Newer variants like Fable 5.1 and Mythos 5 sit just behind at 59–65% depending on the harness, reflecting steady gains from sub-50% levels earlier in the year as models better handle the benchmark’s expert-crafted questions across math, science, and humanities. Competitors including OpenAI’s GPT-6 variants and Google’s Gemini trail on most HLE subsets, though rapid frontier progress leaves room for late-year jumps before the market resolves. Traders weigh these trends against the benchmark’s resistance to saturation and potential protocol differences in final scoring.