Gemini 4 Carbon Now Available: Who Gets It and What It Costs

Gemini 4 Carbon is now on the paid Gemini API, Gemini for business and rolling out to AI Ultra. Prices, the Nov 30 price jump, and when to pick it over Argon.

By PromptWises Editorial Team Published 4 min read
Watch the 1-minute video summary on the PromptWises YouTube channel

PromptWises is reader-supported. Some links may earn us a commission at no extra cost to you. This never affects our ratings or recommendations.

Gemini 4 Carbon, the model Google announced on October 28 for a small group of cyber defenders, became generally available on November 10, 2026. Developers can call it today on the paid tier of the Gemini API and in Google AI Studio, businesses get it in Gemini for business, and it is rolling out to Google AI Ultra subscribers in the Gemini app, AI Mode in Search and Google Antigravity. It costs $2 per million input tokens and $10 per million output tokens until November 30, then doubles. The sources are Google’s announcement by Koray Kavukcuoglu, the developer post and the Gemini API guide to Carbon.

What changed since October

Our first Carbon post covered a model you could not use: access started with cyber defenders in Google’s Fairwind Program, and Google gave no date for anyone else. Two weeks later the general model is out. The cyber-focused version, which Google calls “Gemini 4 Carbon for cyber”, stays inside the Fairwind Program. Google also extended its scientists’ program, which gives selected researchers free access to Gemini 4 Argon in Antigravity for six months, to include Carbon.

Who can use it

WhereWhoStatus
Gemini API, Google AI Studio, Android StudioPaid-tier developers (no free tier)Available, model ID gemini-4-carbon
Gemini for businessEnterprises, via the API or the model drop-downAvailable
Gemini app, AI Mode in Search, Google AntigravityGoogle AI Ultra subscribersRolling out
GitHub CopilotPro+, Max, Business, EnterpriseRolling out gradually
Fairwind ProgramVetted cyber defendersCarbon for cyber

Google’s post names Ultra only for the consumer apps, so AI Pro subscribers should not expect it. GitHub’s changelog says Carbon is billed at provider list price under usage-based billing, and that on Business and Enterprise it is switched on by default unless an admin turned off new models or this one. That is the opposite of Claude Fable 5.2 a day earlier, which Copilot admins had to enable themselves, so admins who care about which models their developers use should check the model policy now.

Pricing

Per million tokens on the paid API, at the time of writing (pricing page):

Through Nov 30, 2026From Dec 1, 2026
Input$2.00$4.00
Output, including thinking tokens$10.00$20.00
Cached input (implicit caching only)$0.10$0.20

The October announcement gave no end date for the introductory price; the November 30 cut-off is now on Google’s pricing page. Carbon and Argon share the same per-token price, but the real bill differs: Google says Carbon “spends more thinking tokens per turn”, so the same request usually costs more and takes longer. Compare that with Fable 5.2 at $8 / $40 per million tokens. Even after December 1, Carbon’s list price is half of Fable’s, although the cost of a finished task depends on how many tokens each model burns, which only your own tests will show.

Carbon or Argon?

Google’s own advice is unusually direct: Argon “remains the ideal choice for the majority of software engineering and knowledge work”. Carbon is for tasks that defeat Argon, and the API guide lists complex root-cause debugging, architectural design, dense multimodal synthesis (large sets of PDFs, spreadsheets and slides), scientific proofs and security auditing.

Technically the two are close. Carbon has a unified 1,048,576-token window shared by input, thinking and output, long decode that lets a single response run past 64K tokens, and three thinking levels: low, medium (the default) and high. Google recommends testing at all three and using the lowest level that meets your quality bar. Carbon also works with the Antigravity agent in Gemini Managed Agents, and Google is previewing a /swarm feature in Antigravity that splits one prompt across a team of agents.

The benchmark claims

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, and second on a new Multimodal Knowledge Agent benchmark built with Artificial Analysis. These are vendor-reported results on benchmarks chosen for the launch. Our guide to reading AI model announcements explains how to check claims like these against your own work.

What to do now

  • Developers: if Argon handles your workload, stay on it. If you have tasks Argon gets wrong, run them on Carbon before November 30 while the price is low, and budget for $4 / $20 after that.
  • Copilot admins: decide whether Carbon should stay enabled under your model policy, since it is on by default.
  • Everyone else: the Gemini review and our Claude vs Gemini comparison cover which plan and assistant to choose. More launches are in the news section.

Frequently asked questions

Can I use Gemini 4 Carbon in the Gemini app?

Only on Google AI Ultra, and it is rolling out, so not every Ultra account has it yet. Google's November 10 post names Ultra subscribers for the Gemini app, AI Mode in Search and Google Antigravity. It does not mention AI Pro or AI Plus.

How much does the Gemini 4 Carbon API cost?

At the time of writing, $2 per 1M input tokens and $10 per 1M output tokens (thinking tokens count as output) through November 30, 2026. From December 1 it is $4 and $20. Cached input is $0.10 per 1M, then $0.20, with implicit caching only. There is no free tier.

Should I use Gemini 4 Carbon or Gemini 4 Argon?

Start with Argon. Google itself says Argon remains the ideal choice for most software engineering and knowledge work. Carbon spends more thinking tokens per turn, so it is slower and costs more per task at the same per-token price. Move a task to Carbon when Argon fails at it: hard debugging, architecture, dense documents, science or security audits.

Is Gemini 4 Carbon available in GitHub Copilot?

Yes, on Copilot Pro+, Max, Business and Enterprise, with a gradual rollout from November 10, 2026. It is billed at provider list price under usage-based billing. On Business and Enterprise it is enabled by default unless an admin turned off the default or this model.

Are the Gemini 4 Carbon benchmark results reliable?

They are reported by Google and by the companies that run the benchmarks with Google, such as Cursor, Perplexity, Zapier and Artificial Analysis. Treat them as a signal, and test Carbon on your own tasks before switching anything important.

Keep reading