Google Opens Gemini 4 Carbon to Paying Users, and Its Launch Price Will Double on December 1

Two weeks ago, Gemini 4 Carbon was the model you could read about but not touch. Google handed it only to cybersecurity defenders in its Fairwind Program, which is why our Oct 29 story ended on a blunt note: you cannot have it. That gate has now opened. Partly.

In a blog post on Tuesday, Google DeepMind SVP and Chief AI Architect Koray Kavukcuoglu said Carbon is now available "to developers, enterprises, scientists, and consumers." The fine print shrinks that list fast. Every route in runs through a paid plan, a subscription at the top of Google's ladder, or an application form.

According to Google, paying developers and enterprises get Carbon through the Gemini API in Google AI Studio and Gemini for business. Google AI Ultra subscribers get it in the Gemini app, AI Mode in Search and Antigravity, the company's agentic coding platform. Scientists can apply for a shot at six months of free access, "subject to eligibility, capacity, and selection." Cyber defenders still go through Fairwind.

Everyone else waits. The Verge reports it is not clear when, or if, lower tier subscribers or free users will get Carbon, and Google's post offers no date.

A price with an expiry date

Then there is the money. Google is pricing Carbon at the same introductory rate as Gemini 4 Argon, its other frontier model: $2 per million input tokens and $10 per million output tokens. A footnote in the post spells out the catch. After the introductory period ends on November 30, the price will be $4 per million input tokens and $20 per million output tokens. Any team wiring Carbon into a product this month is budgeting on a price with a three week shelf life.

Even the doubled rate looks aggressive next to rivals, at least on paper. The Verge reports the introductory price is around a fifth of the cost of OpenAI's GPT-6 Astra and a quarter of Anthropic's Fable 5.2, and that Carbon will still be "significantly cheaper than frontier rivals on a per-token basis" after the increase. The same report adds the asterisk that matters, noting that cheap tokens do not necessarily mean cheap jobs, because models can burn different numbers of tokens on the same task.

That caveat bites harder here than usual. By Google's own description, Carbon spends more tokens than Argon "to explore thoroughly, reason deeply, and run extended verification loops." Google still steers most software engineering and knowledge work to Argon and pitches Carbon at the hard, long jobs: multi-step coding and debugging, and open-ended scientific research. Translation for anyone holding a budget: this is the premium tool, metered by the token, and it is built to use a lot of them.

Google's counterargument is efficiency per job. The company claims Carbon ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench "at more than 2X lower cost per task compared to the next best external model." Those are Google's figures, picked by Google. The Verge notes that companies "can and do cherry-pick benchmark results."

"Argon remains the ideal choice for the majority of software engineering and knowledge work, while Carbon is built for tasks that test the limits of a model's intelligence." Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, Google

The other hook is /swarm, which Google is previewing in Antigravity. The company says it turns a single prompt "into a coordinated team of agents you can track in real time," with Carbon assigned to the heavy reasoning roles such as architecture, coordination and auditing, while faster, lighter models handle work in parallel. It is a preview, not a finished product, and the showcase projects in Google's post are its own demos.

Put the pieces together and the launch looks built to put Carbon in front of the customers most likely to pay, at a price low enough to tempt them for a few weeks. The test comes on December 1, when the discount disappears and developers find out whether the swarm is worth the tokens.