Google Opens Gemini 4 Carbon to All Paid Developers — Its Strongest AI Model Yet, With a Price Doubling on December 1

Google's most powerful AI model is now open to everyone willing to pay for it. On Tuesday, November 10, 2026, Google DeepMind made Gemini 4 Carbon — which it calls its strongest model so far — available to paying developers, enterprises, scientists, and consumers, two weeks after it was first handed only to a small group of cybersecurity defenders. The release also comes with a warning for developers watching their bills: the introductory price doubles on December 1.

The Short Answer

Google opened Gemini 4 Carbon to all paying developers on November 10, 2026. Carbon is the second Gemini 4 frontier model, after Gemini 4 Argon, and Google positions it above Argon for long, multi-step coding, debugging, cybersecurity, and scientific research. It costs $2 per million input tokens and $10 per million output tokens through November 30 — then doubles to $4 and $20 from December 1. It is available through the paid Gemini API in Google AI Studio and Android Studio, in Gemini for business, and for Google AI Ultra subscribers in the Gemini app, AI Mode in Search, and the Antigravity development platform.

From Cyber Defenders Only to Everyone in Two Weeks

Carbon was announced on October 28, 2026, but almost nobody could use it. Google gave it first to a subset of cybersecurity defenders through its Fairwind Program, where it partnered with impacted organizations to remediate vulnerabilities in record time, helping mitigate potential impact on tens of thousands of companies. The November 10 post from Koray Kavukcuoglu, senior vice president of Google DeepMind and Google's chief AI architect, ends that wait: the general-use model is now open to developers, enterprises, scientists, and consumers.

Google describes Carbon as its "strongest model with frontier performance in cybersecurity capabilities and long-horizon agentic workflows." It builds on Gemini 4 Argon but uses more reasoning tokens to investigate problems, evaluate possible solutions, and run additional verification rounds. Google still says Argon is the right pick for most coding and knowledge work — Carbon is for the hardest, longest jobs, and it burns more thinking tokens doing them. The model ID is gemini-4-carbon, with a default thinking level of medium, and it accepts text, image, video, audio, and PDF inputs.

The /swarm Feature: Many Agents, One Prompt

The release adds /swarm, a preview feature in Google's Antigravity environment that can coordinate several coding agents from a single prompt, letting the model build, test, and debug full-stack applications across multiple packages and subsystems. It is Google's answer to the multi-agent coding workflows that have become the frontier of AI-assisted software development — and a direct rival to the agentic coding tools Anthropic and others are shipping. Anthropic launched Claude Fable 5.2 and Mythos 5.2 on Monday, November 9, and Microsoft shipped its PDD Kit for verifying AI-written code the same day.

AI Agents Just Verified Signal's Encryption

On the same day Carbon opened up, Google DeepMind showed what the model can do on hard engineering work. Using a new framework called AlphaProof Code, human experts working with agents running on Carbon — and on Claude Opus 5.5 — turned Signal's security goals into a reviewed Lean specification and produced the machine-checked proofs behind a formal verification of libsignal, the reference implementation of the Signal protocol. DeepMind's announcement summary says three million lines of Lean were involved; SecurityWeek and The Register report about 6.5 million lines. Either way, it is a striking demonstration: AI agents formally verifying the encryption code that protects hundreds of millions of people's messages.

Benchmarks — and the Caveats

Google says Carbon ranks first on CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench, completing tasks at less than half the cost of the next-best model. The published numbers are specific: 53.9 percent on AutomationBench for automation tasks, 88.2 percent on SEC-bench Pro for finding vulnerabilities, 54.8 percent on ExploitGym for checking exploitability, and 71.0 percent on CWE-bench for patching. On the Gray Swan prompt-injection benchmark, Carbon showed a 0.0 percent attack success rate, down from 0.7 percent for Argon.

The caveat is that every one of these figures comes from Google. There are no independent measurements yet, and no model is fully resistant to prompt injection at scale. The pricing deadline is the more concrete fact for developers: the $2 input and $10 output rate ends November 30, 2026, and from December 1 the same tokens cost $4 and $20 — double. Cached input is priced at $0.10 per million tokens. For more on where AI assistants are heading, see our report on Apple's agentic Siri plans for next year. You can follow the latest Google AI and Gemini updates as developers test Carbon in the wild.

Read Also

More Tech & AI coverage from TrendWire Daily: Anthropic Launches Claude Fable 5.2 and Mythos 5.2 and Microsoft's PDD Kit Uses Rust to Verify AI-Written Code.

Frequently Asked Questions

What is Gemini 4 Carbon?

Gemini 4 Carbon is Google's strongest AI model so far, the second frontier model in the Gemini 4 family after Gemini 4 Argon. It is designed for long-horizon agentic workflows — complex, multi-step coding, debugging, cybersecurity, and scientific research.

When did Gemini 4 Carbon become available?

Google announced Carbon on October 28, 2026, and first gave it to cybersecurity defenders through its Fairwind Program. It opened to all paying developers, enterprises, scientists, and consumers on November 10, 2026.

How much does Gemini 4 Carbon cost?

The introductory price is $2 per million input tokens and $10 per million output tokens, the same as Argon. That rate ends November 30, 2026; from December 1 the standard price is $4 per million input tokens and $20 per million output tokens.

Where can I use Gemini 4 Carbon?

Through the paid tier of the Gemini API in Google AI Studio and Android Studio, and in Gemini for business. Google AI Ultra subscribers also get it in Antigravity, the Gemini app, and AI Mode in Search.

What is the /swarm feature?

/swarm is a preview feature in Google's Antigravity development platform that coordinates several coding agents from a single prompt, letting Carbon build, test, and debug full-stack applications across multiple packages and subsystems.

How is Carbon different from Argon?

Carbon sits above Argon in the Gemini 4 family. It uses more reasoning tokens, investigates problems more deeply, and runs more verification rounds. Google says Argon remains the right pick for most coding and knowledge work; Carbon is for the hardest, longest jobs.

Related: Anthropic Launches Claude Fable 5.2 and Mythos 5.2 | Microsoft's PDD Kit Verifies AI-Written Code | Apple's Agentic Siri Coming Next Year