Google has launched Gemini 4 Carbon, a new AI model focused on complex software development and cybersecurity tasks, including detecting vulnerabilities, assessing their exploitability and generating patches.
According to Google, the model is also designed to handle long-term engineering projects that require multiple steps and continuous analysis. The company said its secure-code generation performance was more than twice that of Gemini 4 Argon in internal testing.
Google reported scores of 88.2% on SEC-bench Pro for vulnerability detection, 54.8% on ExploitGym for exploitability assessment and 71% on CWE-bench for vulnerability patching. These results reflect specific testing conditions and may not translate directly to all deployment environments.
Gemini 4 Carbon is initially being offered to selected cybersecurity specialists and organisations through Google’s Fairwind Program. The company has not announced a general availability date.
The model will use the same introductory pricing as Gemini 4 Argon: $2 per million input tokens and $10 per million output tokens.
Source: EntArabi
