Google has released Gemini 4 Carbon, which it calls its strongest model, with first access limited to a group of cyber defenders. The 28 October announcement came two weeks after Gemini 4 Argon. Google says access will widen gradually to trusted partners through its Fairwind Program, and that it will keep refining guardrails before offering Carbon to developers, businesses and consumers. It gave no date for that.
Carbon ranks first on the independent Artificial Analysis Intelligence Index v5.0, with a score of 59. Google reports 88.2% on SEC-bench Pro, which tests finding vulnerabilities, 71.0% on CWE-bench for patching and 54.8% on ExploitGym for exploitation, all measured without guardrails. Other claims, such as more than doubling Argon’s success rate at writing secure code, come from Google’s own testing.
Why it matters: a model this good at finding and exploiting flaws can help attackers as well as defenders, which is why Google is starting with security teams. It says Carbon is part of the US government’s voluntary pre-release model testing process. Introductory pricing is $2 per million input tokens and $10 per million output tokens, rising later to $4 and $20.