Google DeepMind announced Gemini 4 Carbon, positioning it as its strongest model yet for long-horizon engineering and cybersecurity defense work. The company says Carbon scored 59 on the Artificial Analysis Intelligence Index v5, ahead of several rival frontier models.
Because of its offensive-security capability, Google is rolling Carbon out first to vetted cyber defenders through the Fairwind program while it completes voluntary pre-deployment testing with the US government. There is no public date yet for general developer or consumer access.
Google reports strong results on security benchmarks including SEC-bench Pro and CWE-bench, and says pricing is expected to match Gemini 4 Argon at introductory rates. For GenzeRank readers, Carbon is another signal that AI vendors are shipping specialized models for high-risk domains rather than one-size-fits-all chatbots.
Questions? Chat with our team — threaded comments are coming soon.