Google announced on November 10, 2026 that Gemini 4 Carbon is available. It calls Carbon “our strongest model with frontier performance in cybersecurity capabilities”. Two weeks earlier, Carbon was released only to cyber defenders through the Fairwind Program (our October roundup).
What to know
- Who gets it: Google says Carbon is in the API through Google AI Studio and Gemini for business for all paying developers and enterprises. Its developer post says developers start on the paid tier of the Gemini API in AI Studio and Android Studio. For Google AI Ultra subscribers, it is rolling out in Google Antigravity, the Gemini app and AI Mode in Search.
- Price: Google lists an introductory $2 per million input tokens and $10 per million output tokens, the same as Gemini 4 Argon. After November 30, 2026, the public Carbon model goes to $4 and $20.
- Carbon or Argon: Google says Carbon uses more tokens than Argon to reason and check its work, and that “Argon remains the ideal choice for the majority of software engineering and knowledge work.” Carbon is aimed at the hardest tasks, such as complex multi-step coding and open-ended research.
- Swarm in Antigravity (preview): Google says Antigravity’s
/swarmturns one prompt into a coordinated team of agents. Carbon handles architecture, coordination and auditing, and lighter models run tasks in parallel. - Benchmarks (Google’s own figures): Google says 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. It ranks second on the new Multimodal Knowledge Agent benchmark (built with Artificial Analysis), at about 6x lower average cost per task than the top model.
- Safety (Google’s figures): Google says that in simulated cyberattack tests, the per-attempt breach rate fell to 0.8%, from 3.1% on Argon. Carbon passed 92% of a 375-task agentic honesty benchmark, against 70% for Argon. Google says prompt injection robustness “remains an area for continued improvement”.
- Not stated: Google’s posts do not give a preview or general-availability label, a context window or an API model ID. Check the Gemini API developer guide before you build.
Why it matters
The model chooser is now similar at Google and Anthropic: a default tier for most work and a pricier one for the hardest jobs. Compare Carbon with Claude Fable 5.2, and see our guide on choosing between Carbon and Argon. Background on the Gemini page, in the Argon developer summary and in the model tracker.
Source: Google. We summarise in our own words; check the source for the full details.