Google's Gemini 4 series, including variants like 4 Carbon and 4 Argon, has posted strong results on Humanity's Last Exam (HLE), a 2,500-question benchmark of graduate-level expert reasoning across math, physics, biology, and other domains designed to resist simple recall. Recent Artificial Analysis evaluations show Gemini 4 Carbon at 60.4% and Gemini 4 Argon at 57.1% in text-only settings, placing them near or slightly behind Anthropic's leading Claude Opus 5.5 at 61.4%. Tool-augmented runs and verified subsets further highlight competitive positioning, though scores vary by configuration and leaderboard methodology. Google's ongoing model iterations, combined with the benchmark's rolling updates, position Gemini for potential gains before year-end, though saturation effects and rival advances from OpenAI and others create ongoing uncertainty in trader views.