Google unveils Gemini 4 Carbon, its strongest model, and opens it first to cyber defenders
Prepared by Sintelligent with the help of SI. Published in SI: . Updated: .
Google is releasing its strongest model in stages: first to cyber defenders, so that its skills at finding and fixing vulnerabilities reach them before anyone else.
According to Google, Gemini 4 Carbon is "our strongest model with state-of-the-art cybersecurity and long-horizon capabilities." Access is currently limited to a group of cyber defenders; the model will then open to trusted partners through the Fairwind program, and after that to developers, businesses and consumers. Google calls the rollout "deliberate" and says it is taking part in the U.S. government's voluntary pre-release model testing process. According to Google, the model scored 59 on the Artificial Analysis Intelligence Index v5 (first place; according to The Decoder, Gemini 4 Argon scored 50) and placed fourth in WebDev Arena, up 89 Elo points. In tests without safety guardrails, by Google's assessment, it scored 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench; the Chrome team, in a quote provided by Google, rated the quality of its fixes as "10 times better" than that of another lab's large frontier model. The introductory price is $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after the introductory period. According to The Decoder, in agentic coding Carbon scored 59 versus 62 for Claude Opus 5.5.
Sources 2