Market Briefs / ai

Gemini 4 Carbon API prices double on December 1

Google bills Gemini 4 Carbon at an introductory $2 per million input tokens and $10 per million output tokens through November 30. From December 1 the paid-tier rates become $4 and $20, and Gemini 4 Argon moves the same way.

Published

What happened?

Google lists gemini-4-carbon, its most capable reasoning model, on the paid tier of the Gemini API at $2.00 per million input tokens and $10.00 per million output tokens through November 30, 2026. From December 1 the rates are $4.00 and $20.00, and cached input moves from $0.10 to $0.20 (implicit caching only; explicit caching is not supported). gemini-4-argon, the more efficient Gemini 4 model, carries the same rates and the same increase. Neither model is available on the free tier. Google bills every token the model processes, including thinking tokens, which count as output, and the tool instructions and function declarations used with grounding. Grounding with Google Search keeps 5,000 free requests a month, then $14 per 1,000.

Why it matters

The same call costs twice as much from December 1. A job that uses 10 million input tokens and 1 million output tokens costs $30 today and $60 after the change. Because thinking tokens are billed as output, a reasoning-heavy workload can spend more on output than the visible answer suggests, so check real usage instead of estimating. Freelancers who quoted a fixed price, or a retainer that includes API costs, for a project built on Gemini 4 absorb the whole increase unless the quote can change.

What should you do?

Search client projects, environment files and automations for “gemini-4-carbon” and “gemini-4-argon”. Pull last month’s token usage from Google AI Studio and recalculate it at the December 1 rates. Tell clients before December 1 and update quotes or retainers that include API costs. Test whether Gemini 4 Argon, which Google says uses fewer tokens per turn, gives acceptable results on the same prompts. Set a max output token limit and a budget alert, and where the client owns the account, have the API key billed to them directly so cost changes pass through.

Original sources