Google announces Gemini 4 Carbon, not yet available to users
Original title: Google announces Gemini 4 Carbon, its latest frontier AI model that you can't use
Google says Carbon tops Argon on its tests, including cleaner code, but the staged release remains closed.
Two weeks after Gemini 4 Argon, Google announced Gemini 4 Carbon as an improvement rather than the expected Flash-class model. Google says it is stronger at coding, agentic tasks, scientific analysis and cybersecurity work such as vulnerability detection, validation and remediation, and twice as likely as Argon to generate vulnerability-free code. The Chrome team claims Carbon is 10 times better at high-quality patches than non-Google models. Users cannot use it yet. Ars notes rival frontier models are only a few points behind and Carbon might be a post-training upgrade of Argon.
Unconfirmed: Benchmark claims are Google’s; Ars says Carbon may be a post-training upgrade of Argon.
- Google says Carbon’s WebDev Arena score rose 89 points, ranking fourth.
- Google says it beats GPT-6 Astra, Claude Opus 5.5 and Claude Fable 5.1 on almost every published test.
- Google describes a staged release that could take a while, as with Argon.
- Vendors
From the original article
1 min read at arstechnica.com
Gemini 4 Carbon is essentially a slightly better version of Argon.
By Ryan Whitwam · arstechnica.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.