Gemini 4 Carbon targets cyber defense, rolling out to defenders first
Gemini 4 Carbon scores 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench, can run for days to root-cause and patch bugs, and ships first to cyber defenders via the Fairwind Program at Argon's price.
🔥🔥 Google's strongest model yet is built for cyber defense first. Gemini 4 Carbon leads on finding, exploiting, and patching real vulnerabilities: 88.2 percent on SEC-bench Pro, 54.8 percent on ExploitGym, and 71.0 percent on CWE-bench. It is built for long jobs too, running for days to root-cause and patch bugs across large codebases. And it is harder to trick: prompt injection resistance hit 0.0 percent, and cheating on Terminal-Bench fell to 0.0 percent. What makes this valuable: it does not stop at flagging flaws, it turns validated findings into working fixes, and it generates secure code at over 2x the rate of Argon. It is rolling out first to cyber defenders this week through the Fairwind Program, then to developers and businesses. Here is what people can build on it: security pipelines that find, validate, and patch vulnerabilities automatically, coding agents that work multi-day shifts on hard bugs, and web apps with noticeably better output. Same price as Argon when it reaches everyone.