Google has introduced Gemini 4 Carbon, a new frontier model, according to Engadget. The model builds on Gemini 4 Argon, which Google says launched two weeks ago.
Performance claims from the provider
Google states that Gemini 4 Carbon achieves the highest score on the AA Index v5.0, a general AI benchmark, with 59 points – just ahead of Claude Opus 5.5 and GPT-6 Astra. Google also claims a near sweep in categories including agentic coding, computer use, knowledge work, science and math. Cybersecurity is especially highlighted: the model is said to find and prioritize vulnerabilities better and to patch their root causes. According to the company, it is also twice as likely to generate secure code as Argon.
Safety and access
Google further says the model reduces shortcuts and loophole exploitation, improves transparency and honesty (including a test for unsupported claims), and resists prompt injection. These statements come solely from the provider.
Initially, Carbon will only be released to select cybersecurity defense companies to allow pre-release testing. Additional trusted partners are to gain access later through the Fairwind Program before a broad release. Google followed this path with Gemini 4 Argon, which is now available in the Gemini app and AI search mode.
Classification
For development and security teams in Germany, the key question is whether the coding and security claims hold up in independent tests. Until then, they remain vendor claims. It is also open when and to what extent Carbon will be available to companies outside the partner group.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
