Google introduced Gemini 4 Carbon, a model focused on cybersecurity and long-running tasks, with access initially limited to selected defenders and partners.

Google unveiled Gemini 4 Carbon on Oct. 28, 2026, positioning it as a frontier model for cybersecurity defense and long-running engineering work. The company says it scored 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for testing exploitability and 71.0% on Collinear’s CWE-bench for patching flaws. Google also reports more than twice Argon’s success rate in generating secure code and says its Chrome team saw Carbon produce high-quality patches at over 10 times the rate of a rival lab’s model. Google says internal teams have run it for days on complex tasks across large codebases. Access begins with selected cyber defenders through the Fairwind Program, then is expected to expand to vetted partners; general availability has no announced date. That staged release reflects the model’s dual-use capabilities: defenders get tools to find, validate and fix flaws, while broad access could carry greater risk.