Google announces Gemini 4 Carbon, a model not yet available to users

The announcement emphasizes cybersecurity capabilities, with Google claiming Carbon shows a "significant leap" in vulnerability detection and remediation. The company reports that Carbon is twice as likely to generate code free of vulnerabilities compared to Argon. Additionally, the Chrome team claims the model is 10 times better at generating high-quality patches than non-Google models. On the WebDev Arena leaderboard, Carbon has increased its score by 89 points, placing it at number four.

While Google positions Carbon as a distinct new model, it may be a post-training upgrade of Argon, given their similarities. In benchmarks, Carbon outperforms GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1 in most tests, though the margins are slim. The article notes that all frontier models currently score in a similar range, suggesting the competitive gap remains narrow despite the reported improvements.