Google's Gemini lineup has traded leadership on Humanity’s Last Exam (HLE), a 2,500-question expert-level benchmark from the Center for AI Safety and Scale AI, with Gemini 3 Pro and Deep Think posting record scores around 37–48% in late 2025. As of October 13, 2026, Anthropic’s Claude Fable 5.1 and Opus 5.5 variants lead most HLE and AA-HLE leaderboards at 61–65%, while Google’s newly announced Gemini 4 Argon sits at 57.1% and Gemini 3.8 Flash tops certain verified subsets near 55%. The just-released Gemini 4 Argon reflects continued scaling and reasoning gains, yet competitive pressure from Claude and OpenAI models keeps the frontier tight. Traders watch for official model releases or benchmark reports that could shift implied probabilities on whether the next Gemini Pro variant debuts with a leading HLE result.