Google announces Gemini 4 Carbon, its latest frontier AI model that you can't use

Google announced Gemini 4 Carbon, a new frontier AI model arriving two weeks after Gemini 4 Argon. Carbon is not the lower-cost Flash-class model some users expected; Google is staging its release and has not made the model available for general use. Google says Gemini 4 Carbon improves coding, agentic-task performance, scientific analysis, and cybersecurity work. The company reports a significant gain in vulnerability detection, validation, and remediation, and says Carbon is twice as likely as Argon to generate code without vulnerabilities. Google’s Chrome team says Carbon generates high-quality patches at a rate 10 times higher than non-Google models. Google also says Carbon gained 89 points on WebDev Arena and now ranks fourth on that leaderboard. Google presents Carbon as a new model and reports that it exceeds Argon by narrow margins across its published tests. In Google’s benchmarks, Carbon also outperforms GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1 in nearly every test. The reported differences between Carbon and competing frontier models are only a few points, and the models score in a similar range across the cited evaluations.