The American Tech company Google has announced Gemini 4 Carbon, its newest frontier model engineered specifically for advanced cybersecurity defence and sustained, long-horizon agentic workflows. Coming two weeks after the release of Gemini 4 Argon, the new release places a heavy focus on vulnerability detection, risk validation through exploitation testing, and autonomous root-cause patching across complex codebases.
According to Google, Carbon ranks first on the Artificial Analysis Intelligence Index v5.0 and demonstrates significant benchmark gains across coding, penetration testing, and prompt-injection resistance. Alongside security capabilities, the company reported measurable improvements in alignment, citing reduced fabrication rates and fewer attempts to exploit shortcuts during task execution. Rollout is taking a phased approach, beginning with select cybersecurity defenders and trusted partners before expanding to general developers, businesses, and consumers.
