Gemini 4 Carbon: Features, Pricing and How to Access It

What is Gemini 4 Carbon?

Gemini 4 Carbon is Google DeepMind's most capable AI model, built to handle demanding technical workflows that test the limits of machine reasoning. Google released the model publicly on 10 November 2026 after providing two weeks of early access to security teams in its Fairwind Program.

Google positions Carbon above its recent Gemini 4 Argon release. While Argon remains Google's primary recommendation for everyday coding and office tasks, Carbon consumes more tokens to explore options, run verification loops, and resolve multi-step problems.

How to access Gemini 4 Carbon

Google provides access to Gemini 4 Carbon across developer platforms, enterprise plans, and consumer subscriptions. The model is not currently available on free tiers or standard consumer plans.

• Consumer subscribers: Google AI Ultra members can access Gemini 4 Carbon inside the Gemini web and mobile apps, Google Antigravity, and AI Mode in Google Search.
• Developers and businesses: Paying developers can access the model via the Gemini API in Google AI Studio, as well as through Gemini for business subscriptions.
• Scientists and researchers: Researchers can apply to Google for up to six months of complimentary access to both Carbon and Argon.
• Cybersecurity teams: Security professionals can apply for specialized access through Google's Fairwind Program.

How agent swarms work in Google Antigravity

Google introduced a preview command called /swarm inside its Google Antigravity development environment. The feature turns a single user prompt into a coordinated team of AI agents that execute complex projects in parallel.

In a swarm workflow, Gemini 4 Carbon takes on the roles that require the deepest reasoning, including architectural design, multi-agent coordination, and code auditing. Faster and lighter models handle isolated subtasks simultaneously while adversarial verifiers evaluate each change.

In testing highlighted by Google DeepMind, Carbon agent teams built interactive Three.js 3D web applications, synthesized architectural planning documents, and assembled a playable browser game from single prompts.

What Gemini 4 Carbon can do

Gemini 4 Carbon specializes in complex software engineering, cybersecurity research, and multimodal data extraction. It runs iterative build-and-verify loops during software development, inspecting rendered user interfaces and fixing code errors automatically.

For enterprise operations, the model processes large collections of dense documents containing schematics, tables, and blueprints. In financial evaluations, Carbon matched top-tier models on Hebbia's E2E Financial Agentic benchmark and ranked first out of 23 configurations on Balyasny Asset Management's evaluation.

In cybersecurity tests, security firm Wiz used the model to discover zero-day vulnerabilities across two major SaaS providers in less than two hours. In laboratory applications, drug toxicity researcher Axiom Bio used Carbon to detect previously unknown toxicity mechanisms.

Gemini 4 Carbon pricing and benchmarks

Google launched Gemini 4 Carbon at an introductory API rate of $2 per million input tokens and $10 per million output tokens. According to The Verge, these introductory token rates will double at the end of November 2026.

On external benchmarks, Carbon secured first place on CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench at less than half the per-task cost of rival models. On the Multimodal Knowledge Agent benchmark developed with Artificial Analysis, Carbon ranked second while operating at more than six times lower cost per task than the highest-performing model.

Frequently asked questions

Sources: Google · Digital Trends · The Verge