Anthropic’s May 28 release of Claude Opus 4.8, featuring adaptive reasoning and max-effort modes, propelled the model to a leading 45.7% on Humanity’s Last Exam, a 2,500-question expert benchmark spanning math, physics, biology, and humanities designed to resist saturation. This edges out Gemini 3.1 Pro Preview at 44.7% and recent GPT-5.5 variants near 44.3%, building on earlier Claude Opus 4.7 and 4.6 thinking-mode results in the mid-30s. With the June 30 resolution deadline weeks away, traders are monitoring potential new Claude iterations, inference optimizations, or leaderboard updates that could extend the narrow lead before the market closes. Human experts average near 90%, underscoring the benchmark’s difficulty for frontier large language models.