Anthropic’s April 2026 releases of Claude Opus 4.7 and the Mythos Preview have lifted frontier Claude models to leading scores of 35–65% on Humanity’s Last Exam, a 2,500-question benchmark of expert-level reasoning across math, science, and humanities. These gains stem from scaled oversight techniques, extended chain-of-thought prompting, and tool integration that outperform earlier Claude versions while edging OpenAI’s GPT-5 series and Google’s Gemini 3.1 Pro on the public leaderboard. With June 30 just weeks away, trader focus centers on whether a new public Claude variant or benchmark update will push any model past the specific resolution thresholds before competitive releases or evaluation changes alter the picture.