Google's Gemini 4 Carbon security benchmark claims versus rival models

Google's Chrome team says Gemini 4 Carbon generated more than 10 times as many high-quality security patches as another lab's large frontier model, according to a post by Paige Bailey that quotes the team.

Google also reports more than 2 times higher secure-code success than Argon, nearly 80% fewer high-severity misalignment alerts, and zero successful attacks on its evaluated prompt-injection benchmark. Bailey's post also cites an Artificial Analysis score of 59 for Carbon against 58 for Claude Opus 5.5. These are figures Google reports for its own evaluations; no independent verification is reported.