Gemini 4 Carbon is a frontier artificial intelligence model announced by Google DeepMind on 28 October 2026. Google describes it as its most capable model, with gains in cyber defence and long-horizon task execution.
Release and availability
Carbon uses the same safety and safeguards stack as Gemini 4 Argon, released on 30 September. At launch it was available only to a small group of vetted cyber defenders; Google plans to widen access through the Fairwind Program to more trusted partners and then to developers, enterprises and consumers. Google says it is taking part in the US government's voluntary pre-deployment testing process for frontier models.
Evaluation
Carbon ranks first on the Artificial Analysis Intelligence Index v5.0 with a score of 59, and fourth on WebDev Arena with 1,768 points. Google's cybersecurity results, measured with guardrails disabled, were 88.2% on SEC-bench Pro for vulnerability discovery, 54.8% on ExploitGym for exploitability validation and 71.0% on CWE-bench (Collinear) for root-cause remediation. Google also reports that it generates secure code more than twice as often as Argon, cuts high-severity misalignment alerts by nearly 80%, and reduces the success rate of Gray Swan indirect prompt injection attacks to 0.0%.
Pricing
Introductory pricing matches Argon at US$2 per million input tokens and US$10 per million output tokens, with cached input billed at a 95% discount; after the introductory period the prices are US$4 and US$20 per million tokens.