Google debuts Gemini 4 Carbon, a top-ranked AI model still unavailable to the public
The new system outperforms rivals in coding and security tests, but Google has not yet opened it to general users.
By Mei Chen ·

Google announced Gemini 4 Carbon on Tuesday, its newest frontier artificial intelligence model. The release follows Gemini 4 Argon, which the company launched two weeks ago. Carbon is not a lightweight Flash version, but an advanced update to the Argon architecture.
The company claims Carbon reaches a new threshold in coding, agentic tasks, and scientific analysis. Google stated the model shows a significant leap in cybersecurity capabilities, specifically in vulnerability detection and remediation. Internal data indicates Carbon is twice as likely to generate code without vulnerabilities compared to Argon.
The Chrome engineering team reported specific performance gains. They claim Carbon generates high-quality patches 10 times better than non-Google models. Carbon also improved its score by 89 points on the WebDev Arena, moving it to the number four position on that leaderboard.
Google positions Carbon as a distinct new model, though it may be a post-training upgrade of Argon. The two systems share many similarities, but Carbon edges out Argon by small margins in every test provided. In head-to-head comparisons, Carbon beats GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1 in almost every category.
Benchmarks alone do not provide a complete picture of the model's capabilities. All current frontier models score in similar ranges on these tests, with gaps between competitors often amounting to just a few points. Google has not yet opened Carbon to the public, continuing its staged release strategy that may take considerable time to complete.






