Google Debuts Gemini 4 Carbon AI Model Targeting Cybersecurity Defenses
Alphabet’s Google unveiled Gemini 4 Carbon on Wednesday, a frontier artificial intelligence model engineered for advanced cybersecurity tasks, secure code generation, and complex software patching. Software engineers and cybersecurity defenders gain access to a system designed to handle long-running engineering operations as Google expands its model family just two weeks after introducing Gemini 4 Argon.
Early Access Limited to Fairwind Program Participants
The Carbon model delivers state-of-the-art performance across vulnerability discovery, exploitation, and software patching, generating secure code at more than twice the success rate of Gemini 4 Argon on internal benchmarks. Google has initially restricted availability to a subset of cybersecurity defenders participating in the Fairwind Program while the corporation continues testing operational safeguards.
The system currently holds the top spot on SEC-bench Pro, a benchmark measuring an artificial intelligence model’s ability to discover real-world bugs, alongside ExploitGym, which measures exploit generation from vulnerabilities. On the Artificial Analysis Intelligence Index, Carbon outranks competing systems including Claude Opus 5.5 and GPT-6 Astra.
Gemini 4 Carbon is priced at $2 per million input tokens and $10 per million output tokens, matching the introductory pricing structure established for Gemini 4 Argon.
