Gemini 4 Carbon: our next frontier model

Google announced Gemini 4 Carbon, its next frontier model and strongest to date, featuring state-of-the-art cybersecurity and long-horizon capabilities. It ranks #1 on the Artificial Analysis Intelligence Index v5.0 with a score of 59 and shows major gains in coding and agentic tasks, including leading results on SEC-bench Pro (88.2%) for vulnerability discovery, ExploitGym (54.8%) for exploitation, and CWE-bench (71.0%) for patching. Carbon is built for sustained agentic workflows, capable of running for days on complex engineering tasks, and reaches #4 on WebDev Arena with 1769 elo. On safety, it reduces cheating attempts on Terminal-Bench 4.0 to near zero, cuts high-severity misalignment alerts by almost 80%, halves unsupported claims versus Argon, and achieves a 0.0% prompt-injection attack success rate on the Gray Swan benchmark. Following a deliberate rollout, Carbon will first be available to a subset of cyber defenders and trusted partners through the Fairwind Program, then to developers, businesses, and consumers. It launches at the same introductory price as Argon: $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95% off.

(Source：Gemini)