Google announced Gemini 4 Carbon today, its newest frontier model arriving just two weeks after Gemini 4 Argon. The company calls Carbon its strongest model yet, built specifically for sustained agentic workflows and long-horizon engineering problems. Internally, Google teams have watched it run for days at a time, persisting through dead ends to root-cause and patch subtle bugs across large codebases and low-level systems.
The headline number is Carbon's rank on the updated v5.0 Artificial Analysis Intelligence Index, where it scores 59. That is one point ahead of Claude Opus 5.5 and well clear of Argon's 50. Google also reports a firm lead in SEC-bench Pro at 88.2 percent for finding vulnerabilities, ExploitGym at 54.8 percent for exploitation, and CWE-bench at 71.0 percent for patching. In web development, Carbon jumped 89 elo points to land fourth on WebDev Arena.
Access is deliberately narrow. Starting this week, Carbon goes to a small group of cybersecurity defenders through Google's Fairwind Program, with more trusted partners phased in gradually before any rollout to developers, businesses, or consumers. Pricing matches Argon's introductory rate, with cached input tokens discounted 95 percent, while a higher rate takes effect once the introductory period ends. Google says it is taking part in the United States government's voluntary pre-release model testing process.