Gemini 4 Carbon

Google announced on 10 November 2026 that Gemini 4 Carbon is now available to developers, enterprises, scientists, and consumers, following its two-week Fairwind Program release to cybersecurity defenders. Google positions Argon as the default choice for most software engineering and knowledge work, with Carbon aimed at complex multi-step coding and debugging, open-ended scientific research, and extended verification loops that consume more tokens. Carbon ranks first on CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench, at more than 2X lower cost per task than the next best external model. Google is also previewing /swarm in Google Antigravity, turning one prompt into a coordinated team of agents that can be assigned roles such as architectural design, coordination, and auditing, with the model available via API in AI Studio and Gemini for business at the same introductory price as Argon.

Google announced Gemini 4 Carbon on 28 October 2026 as its next frontier model, calling it its strongest model with state-of-the-art cybersecurity and long-horizon capabilities. It builds on Gemini 4 Argon, which launched two weeks earlier, and the company says it is engaged in the U.S. government's voluntary pre-release model testing process while access starts with cyber defenders through the Fairwind Program. Carbon ranks first on Artificial Analysis Intelligence Index v5.0 with a score of 59, and Google reports 88.2% on SEC-bench Pro for vulnerability finding, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching, alongside internal gains in secure code generation. Launch pricing matches Argon at $2 per million input tokens and $10 per million output tokens, with cached input priced at 95% off the input rate.

Google DeepMind's Gemini 4 Carbon model evaluation document details the methodology behind the model's published results. Gemini 4 Carbon and Gemini 4 Argon scores are pass@1 with no majority voting or parallel test-time compute, run through the Gemini API at the highest thinking settings unless noted, and averaged over multiple trials for smaller benchmarks to reduce variance. The evaluation spans agentic coding, computer use, knowledge work, science and math, multimodal capabilities, long-context, cybersecurity, and the Artificial Analysis Intelligence Index, with benchmarks as of October 2026. Some results are self-computed with a mini-swe-agent harness for SWE-sweep and the 122-task Agents' Last Exam v1.1 with the ALE-Claw harness, while non-Gemini results rely on provider self-reporting or public leaderboards.

Google DeepMind's model card for Gemini 4 Carbon, published 10 November 2026, describes the model as having frontier performance in long-horizon agentic workflows across complex knowledge work, engineering, cybersecurity and scientific reasoning. It is based on Gemini 4 Argon, with architecture, training dataset, data processing, hardware and software details delegated to the Argon model card. Carbon accepts text strings, images, audio and video files as input, and returns text output, with a combined input-output token context window of up to 1M tokens. Distribution is listed across the Gemini App, Gemini API, Google AI Studio, Gemini for business, Google Antigravity, Google AI Mode, and the Fairwind Program for verified cybersecurity defenders, with no required hardware or software to use it.