Recent model releases from Anthropic, OpenAI, and Moonshot AI have driven the highest reported Humanity's Last Exam scores to 59–65% without tools and up to 70% with tool use as of early November 2026, up from roughly 9% in early 2025. Claude Fable 5.1 and Opus 5.5 variants currently lead several leaderboards on the 2,500-question benchmark, while GPT-6 Astra tops the HLE-Diamond subset at around 60%. Competitive pressure among frontier labs, continued scaling of large language models, and agentic tool integration explain the rapid gains. Key upcoming catalysts include further 2026 model launches, developer conferences, and potential benchmark refinements that could shift the top score before year-end. Trader consensus reflects these verified capability jumps rather than guarantees of saturation.