Google's Gemini models currently trail the Humanity's Last Exam (HLE) leaderboard, with recent versions like Gemini 3.1 Pro and 3.8 Flash posting scores in the mid-to-high 40s percent on text-only or no-tools evaluations, compared to Anthropic's leading Claude Opus 5.5 and Fable 5.1 near 60-65%. HLE's 2,500 expert-authored questions test frontier multi-step reasoning across disciplines, where progress has accelerated from single digits in early 2025 to the current range through improved large language model capabilities. Competitive dynamics favor rapid iteration by Google DeepMind on model scale, training techniques, and post-training alignment, though no confirmed Gemini release has yet closed the gap to the top. Key upcoming catalysts include potential late-2026 Gemini updates or developer conferences that could demonstrate gains on verified HLE subsets before the December 31 resolution. Trader sentiment reflects uncertainty over whether Google's next iterations can match the pace of Anthropic and OpenAI advances on this expert-level benchmark.