FreshSynapse

AI trends & daily updates, in plain English

AI Models

Google's Gemini 4 Carbon Is Out of the Lab and Into the Gemini App — for Ultra Subscribers

By FreshSynapse Editorial · · 4 min read

A row of tall black mesh-front server racks with red cables recedes down a bright data-centre corridor with a pale floor

Photo: Hippopx (public domain). Illustrative image.

What happened. Google made Gemini 4 Carbon available on Tuesday 10 November 2026, thirteen days after first announcing it. Paying developers can use it through the Gemini API in AI Studio and Android Studio, enterprises through Gemini for business, and Google AI Ultra subscribers are getting it in the Gemini app, AI Mode in Search and Google Antigravity. Carbon is the model Google calls its strongest, built for long-horizon agent work, cybersecurity and scientific reasoning — and until now it had been limited to a small group of cybersecurity defenders in a programme called Fairwind.

The price has a calendar attached. The introductory API price is $2 per million input tokens and $10 per million output tokens — the same as Gemini 4 Argon — and every processed token is billed, including thinking tokens. That rate ends on 30 November 2026; from 1 December the prices double to $4 and $20. Carbon is excluded from the Gemini API free tier. For everyday users the catch is the plan: Google’s own availability table lists Carbon only on the Google AI Ultra row, and Ultra starts at $99.99 a month in the US. Free Gemini, Google AI Plus and Google AI Pro subscribers are not named in the launch post, and Google’s plans page lists Pro as including Argon, not Carbon. The specialised cybersecurity configuration of Carbon remains gated through Fairwind applications on a rolling basis.

What Carbon is for. Carbon builds on Argon’s pre-training foundation and architecture, with additional post-training for deep reasoning, extended verification and multi-step agent workflows, according to the model card. The model ID is gemini-4-carbon, with a 1-million-token window shared by input, thinking and output. The model card’s results, current as of October 2026, report 59 for Carbon on Artificial Analysis’s Intelligence Index v5.0 against 50 for Argon, and 96.2% against 84.2% on the GraphWalks long-context test from 256,000 tokens to 1 million. The listed knowledge cutoff is March 2026, and the card names hallucinations, possible slowness or timeouts, and greater token consumption at higher thinking levels as limitations. Google says its safety report found Carbon did not cross its material-capability-change threshold relative to Argon, so no full new critical-capability assessment was required. Google also says Argon remains its recommended option for most software-engineering and knowledge-work tasks — Carbon uses more tokens to explore and verify harder problems.

The other half of the launch. Google is rolling out /swarm in preview in Antigravity, its coding tool — a feature for running teams of agents together. And it extended its programme giving scientists free access: selected current and future participants now get both Argon and Carbon in Antigravity’s highest usage tier free for six months, with applications open and rolling onboarding subject to eligibility, capacity and selection.

Why it matters. Carbon’s move from a restricted security preview to a paid product in thirteen days is the same thirteen-day pattern Argon followed — announced 30 September, in the app 13 October — and it says something about how Google now ships its frontier models: announce to defenders first, then open the till. The pricing is the real story, though. At $2/$10 until the end of the month, Carbon undercuts Claude Fable 5.2’s $8/$40 by a wide margin for the same long-horizon work; at $4/$20 from December, the gap narrows and the token-hungry design matters more. For developers the question is whether Carbon’s extra verification is worth the extra tokens. For everyone else, the question is simpler: Google’s strongest model now costs $99.99 a month to touch, and the free tier is not invited.

Sources

  • Google
  • Gemini 4 Carbon
  • AI models
  • Google AI Ultra

Related articles