Gemini 4 Carbon: our next frontier model

Open original

Google is releasing Gemini 4 Carbon, a frontier model it says leads on coding, agentic work, and cybersecurity tasks like finding, exploiting, and patching vulnerabilities. Access is rolling out first to a small group of cyber defenders and trusted partners before a broader release.

  • Carbon scores 88.2% on SEC-bench Pro for finding real vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching, which Google describes as state-of-the-art.
  • On internal tests it cuts unsupported claims by about 2x versus Gemini 4 Argon, and it resisted every prompt-injection attempt in Gray Swan's benchmark, with a 0.0% attack success rate.
  • Rollout is gated: it goes to selected cyber defenders first, then trusted partners through the Fairwind Program, with general developer and consumer access to follow once guardrails are refined.

Google's Gemini 4 Carbon is being positioned as its strongest model yet, with the headline claim being a #1 ranking at 59 on the Artificial Analysis Intelligence Index (v5.0). The bigger emphasis, though, is cybersecurity. Carbon scores 88.2% on SEC-bench Pro for finding real vulnerabilities, 54.8% on ExploitGym for checking whether a flaw is actually exploitable, and 71.0% on CWE-bench for writing patches that fix root causes. Google says Carbon produces more than twice the success rate of its predecessor Argon on secure code generation, and that the Chrome team found it outperformed another lab's large frontier model by more than 10x on high-quality patches. Coding gains are also highlighted, including multi-day agentic runs on root-causing bugs across large codebases, and a jump of 89 Elo points on WebDev Arena to rank #4. The safety section is where the claims get more interesting, and also where you should read with some skepticism. Google reports that successful cheating on Terminal-Bench 4.0 fell to 0.0%, that high-severity misalignment alerts in internal live coding dropped by almost 80%, and that unsupported claims were cut roughly in half compared with Argon. It also says overreach fell 26% and under-service fell 29%, and that Carbon resisted every prompt injection attempt in Gray Swan's benchmark, down from Argon's 0.7% attack success rate. Most of these numbers come from Google's own internal evaluations, and the benchmark results are described as run without guardrails, so independent verification matters here. On availability, Carbon is rolling out first to a limited group of cyber defenders and trusted partners through the Fairwind Program, with broader release to developers, businesses, and consumers to follow once guardrails are refined. Google is working with the U.S. government's voluntary pre-release testing process. Pricing matches Argon's introductory rate: $2 per million input tokens, $10 per million output tokens, and cached input at 95% off. Watch for the wider launch and any third-party evaluations, since that is when these benchmark claims will actually get tested.

Ask about this article