On 10 November 2026 Google made Gemini 4 Carbon available to all paying developers and businesses. When Google unveiled the model on 28 October, only selected cyber defenders in its Fairwind programme could use it. Now it is on the paid tier of the Gemini API in Google AI Studio, in Android Studio and in Gemini for business. Google AI Ultra subscribers also get it in the Gemini app, AI Mode in Search and Google Antigravity. The cyber-focused version is still Fairwind-only.
Price: during an introductory period Carbon costs $2 per million input tokens and $10 per million output tokens, the same as Gemini 4 Argon. After 30 November 2026 it goes up to $4 and $20. VentureBeat notes that this matches Claude Opus 5.5 ($4/$20) and is half of Claude Fable 5.2 ($8/$40).
Google's claims: first place on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench at "more than 2X lower cost per task" than the next best outside model. VentureBeat found no published Carbon scores for CursorBench or WANDR and treats those claims as unverified. In Artificial Analysis's index, Carbon (high) scores 60, behind Fable 5.2 (65) and ahead of Opus 5.5 (59). On coding tests such as FrontierCode 2.0 and FrontierSWE v2, Fable 5.2 leads.
Limits: Google itself says Carbon uses more tokens than Argon and that Argon remains its recommended choice for most software engineering work. A lower token price does not always mean a cheaper task, VentureBeat warns. Google also says resistance to prompt injection "remains an area for continued improvement". Vertex AI availability and a country list were not given.
