Google made Gemini 4 Carbon broadly available on 10 November 2026, according to Google’s own announcement. The model had been introduced on 28 October 2026 with access limited to a subset of cyber defenders through the Fairwind Program; the 10 November post extends it to paying developers and enterprises and begins a rollout to Google AI Ultra subscribers.
What happened
On 28 October 2026 Google introduced Gemini 4 Carbon as its next frontier model, with initial availability restricted to a subset of cyber defenders and gradual expansion to trusted partners through the Fairwind Program. Google said it would bring the model to developers, businesses and consumers later, without a date. On 10 November 2026 Google published a follow-up post, “Gemini 4 Carbon is now available,” saying the model is available to paying developers and enterprises and is rolling out to all Google AI Ultra subscribers.
Key details
- Access tiers. Per Google’s 10 November post: paying developers can use the Gemini API paid tier in Google AI Studio and Android Studio; enterprises can use the API or the model selector in Gemini for business, plus Google Antigravity or CodeMender; Google AI Ultra subscribers are receiving access in Google Antigravity, the Gemini app and AI Mode in Search; cybersecurity defenders can use a Carbon variant for cyber through the Fairwind Program on a rolling basis.
- Researcher access programme. Google says applicants may receive free access to Gemini 4 Argon and Carbon through Antigravity’s highest usage tier for six months, onboarded on a rolling basis and “subject to eligibility, capacity, and selection.”
- Pricing. Introductory API pricing is $2 per million input tokens and $10 per million output tokens, the same as Gemini 4 Argon. Google states that after the introductory period ends on 30 November 2026, the price rises to $4 per million input tokens and $20 per million output tokens. A footnote says this applies to the Carbon model intended for public access. Google’s 28 October post also states that cached input tokens cost 95% less than standard input tokens.
- Safety framing. Google says Carbon uses the same multi-layered safety stack as Argon and that it is participating in the US government’s voluntary pre-release model testing. Benchmark and safety results in the announcement are Google’s own reports.
- Data terms. The announcement posts do not describe data-handling, retention or training terms for Carbon. Those are set by the applicable Gemini API, Google Cloud or Workspace terms for each product and tier, which institutions should read directly.
What it means for researchers and research offices
- The introductory price is time-limited. Any cost estimate built on $2/$10 per million tokens should be re-run at $4/$20, which is double, since the stated introductory period ends 30 November 2026. See our note on AI token costs and institutional budget planning.
- Access tier determines terms. Consumer subscriptions (AI Ultra), the paid developer API and enterprise offerings are different products. Procurement and research-data governance staff should confirm which data-use terms govern the tier a lab actually uses before sensitive or unpublished data is submitted.
- The researcher programme is selective. Free six-month access is by application and subject to capacity; it is not a general entitlement, and grant budgets should not assume it.
- Model churn. Argon, then Carbon, within about two weeks illustrates the release cadence that policy reviews must accommodate; see AI model release cadence and institutional policy review. Methods sections and AI-use disclosures should record the exact model name and access date.
What to watch
- Whether the $4/$20 price takes effect on 30 November as stated.
- Published data-processing and retention terms for Carbon on the API and Workspace tiers.
- Independent evaluations of Carbon’s reported benchmark results.
- Other announcements from Google about additional developer features; this article covers only what Google’s Carbon posts state.
