Google's recent Gemini 4 releases, including Argon in September and Carbon in late October 2026, have already posted competitive self-reported scores on refined Humanity's Last Exam variants such as HLE-Diamond and HLE-Verified, where models like Gemini 4 Carbon and 3.8 Flash rank near the top alongside Claude Opus variants. The September 2026 launch of HLE-Diamond, a cleaned 1,000-question subset emphasizing expert reasoning and knowledge, has sharpened focus on frontier benchmarks that resist saturation. With Gemini 4 Hydrogen arriving November 12, the next Pro-level iteration faces pressure from Anthropic's strong Claude Fable results and OpenAI's GPT-6 entries; traders watch for official announcements, API releases, or independent evaluations that could signal Google's strategy for model debuts amid tight competitive timelines.