Google has released Gemini 4 Carbon to developers and enterprises through the paid API in Google AI Studio and Gemini for Business, and is rolling it out to Google AI Ultra subscribers in the Gemini app, Antigravity and AI Mode in Search. The model is described as the most capable in the Gemini 4 family, focused on deep reasoning, long-horizon agents and cybersecurity.
What happened?
Announced on November 10, 2026, Gemini 4 Carbon (model ID gemini-4-carbon) arrives with a context window of up to 1 million tokens. Introductory pricing is $2 per million input tokens and $10 per million output tokens — the same as Gemini 4 Argon — valid until November 30, 2026; after that it rises to $4 and $20. Free-tier and AI Pro users do not have direct access.
Two weeks earlier the model had been limited to cybersecurity defenders in the Fairwind Program. It is now open to paying API users and Ultra subscribers. Google highlights top performance on benchmarks such as CursorBench 4.0, WANDR and AutomationBench, with lower cost per task than competitors.
Why does it matter?
Carbon is positioned for tasks that demand extended reasoning: complex software engineering, scientific research, analysis of dense documents and multi-step agents. In Antigravity, Google is previewing the /swarm feature that turns a single prompt into a coordinated team of agents.
For cybersecurity, Carbon for cyber leads on benchmarks for finding, exploiting and patching vulnerabilities. Google reports internal uses that improved detection of logic flaws in open-source code.
What changes in practice?
Developers can start using the model today in AI Studio (paid tier) and SDKs. Ultra subscribers gain access in the app and agent tools. The introductory price encourages testing through the end of the month. The model handles text, images, audio, video and PDFs, with configurable thinking levels.
Primary source: official Google blog post of November 10, 2026 and DeepMind model card.
Image credit: Google | Source: official blog
By Grok
