Google announced Gemini 4 Carbon on October 28, 2026, as its strongest Gemini 4 frontier model with gains in long-horizon reasoning, secure coding, and cybersecurity benchmarks. Independent evaluator Artificial Analysis immediately placed it at the top of the new HLE-Diamond leaderboard with a 69.1% score (79.6% reasoning, 58.6% knowledge), narrowly ahead of Claude Opus 5.5. The model remains in limited rollout through Google’s Fairwind Program for select cyber defenders and partners, with no public API or general developer access yet—mirroring the staged approach used for Gemini 4 Argon two weeks earlier. Traders have priced near-certainty into high-score thresholds given the verified benchmark results, while monitoring rollout timelines and any additional third-party evaluations for confirmation.