source: Google DeepMind: Gemini 4 Carbon: Our next frontier model
level: technical
google deepmind announced gemini 4 carbon, its strongest model with state-of-the-art cybersecurity and long-horizon capabilities. carbon ranks first on the artificial analysis intelligence index v5.0 with a score of 59. it shows significant gains in coding, agentic tasks, and cybersecurity defense. the model will be available first to a subset of cyber defenders through the fairwind program, with broader access planned later.
carbon achieves 88.2% on sec-bench pro for finding vulnerabilities, 54.8% on exploitgym for exploitation, and 71.0% on cwe-bench for patching. it reduces successful cheating on terminal-bench 4.0 to 0.0% and resists all evaluated prompt injection attacks. internal tests show a 2x uplift in secure code generation over argon and an 80% drop in high-severity misalignment alerts. pricing starts at $2 per million input tokens and $10 per million output tokens.
the model builds on gemini 4 argon, launched two weeks prior, and is part of google's voluntary pre-release testing with the u.s. government. carbon is designed for sustained agentic workflows, running for days on complex engineering tasks. it also improves web development, jumping 89 elo points to rank fourth on webdev arena. google plans to gather feedback from initial users before wider release to developers and consumers.
why it matters: carbon's strong benchmark results in vulnerability detection and secure code generation could help cybersecurity teams automate defense tasks more reliably.
source: Google DeepMind: Gemini 4 Carbon: Our next frontier model