Google has unveiled Gemini 4 Carbon, its latest artificial intelligence (AI) model designed to handle complex coding tasks, long-running engineering projects and cybersecurity challenges, the model offers improved performance in identifying software vulnerabilities, validating security risks and fixing coding issues.

According to an official blog post published by Google yesterday, Gemini 4 Carbon ranked first on version 5.0 of the Artificial Analysis Intelligence Index, scoring 59 points. The model also showed improvements in coding, reasoning and tasks that require AI systems to work through multiple steps over longer periods.

Google said Gemini 4 Carbon achieved an 88.2 percent score on SEC-bench Pro for finding software vulnerabilities, 54.8 percent on ExploitGym for testing whether vulnerabilities can be exploited, and 71 percent on CWE-bench for fixing security weaknesses.

The company said its internal tests showed that Gemini 4 Carbon was more than twice as successful as Gemini 4 Argon at generating secure code. Google also reported improvements in complex software engineering tasks, including finding and fixing bugs across large codebases. Its teams have used the model to work on challenging engineering problems that can take several days to complete. Google is initially making Gemini 4 Carbon available to a selected group of cybersecurity experts.