Google opens Gemini 4 Carbon to paying developers; API rates double to $4 input, $20 output per 1M tokens on Dec. 1

Google opened Gemini 4 Carbon, its newest reasoning model, to paying developers and enterprises on November 10. Introductory API prices expire November 30, and standard input and output rates double on December 1, according to a report from eWeek.

Rollout

The November 10 expansion follows an October 28 rollout that was limited to selected cybersecurity defenders. Carbon is now available through Google's paid developer API and its enterprise products.

In its November 10 announcement, Google positioned Carbon for complex coding, scientific research, cybersecurity and multistep AI workflows.

API pricing

Google's published API rates, per 1M tokens:

Token type	Through Nov. 30	From Dec. 1
Standard input	$2	$4
Standard output	$10	$20
Cached input	$0.10	$0.20

Carbon supports implicit caching.

Carbon and Gemini 4 Argon

eWeek reports that Carbon's per-token prices match those of Gemini 4 Argon. The outlet frames the choice for enterprise AI teams as a cost question: whether Carbon's deeper reasoning justifies the additional computing costs, or whether Gemini 4 Argon can handle the same workloads more economically.

What this means

The price change is scheduled, not a surprise: the introductory rates carry a fixed end date of November 30, and the standard rates take effect the next day. Teams that plan to run Carbon in production need to budget against the December 1 figures of $4 per 1M input tokens and $20 per 1M output tokens, not the introductory ones.

The cached input rate is also part of the change, moving from $0.10 to $0.20 per 1M tokens. Because Carbon supports implicit caching, workloads that resend large shared prompts are affected by that line item as well as by the standard input and output rates.

The practical evaluation, as eWeek describes it, is a workload-by-workload comparison between Carbon and Gemini 4 Argon. Google's targeted use cases for Carbon are complex coding, scientific research, cybersecurity and multistep workflows. Those are the areas where a team would test whether Carbon's reasoning depth justifies its compute costs over Argon.

The source article does not report Carbon's context window, benchmark scores or parameter count.