Price the December run before you switch a coding workflow to Carbon in November.
Google made Gemini 4 Carbon generally available in the Gemini API on November 10. The API release notes list the public model ID as gemini-4-carbon. Google also announced a rollout for Google AI Ultra subscribers through Antigravity. API billing and access through an Ultra subscription are separate.
What the November offer changes
Google lists introductory rates of $2 per million input tokens and $10 per million output tokens through November 30. After the offer, the public model’s standard rates rise to $4 per million input tokens and $20 per million output tokens.
A run with 100,000 standard input tokens and 20,000 output tokens comes to $0.40 at the introductory rates and $0.80 afterward. That estimate uses the published rates for the stated token counts. Cached input and other billing rules can change the actual invoice.
Run a test that includes review time
Start Carbon and Argon from separate copies of the same commit. Pick a bug that crosses a few files and comes with a failing test. Give both models the same tool permissions and stopping rule. Record the tokens used for the full run, then count the minutes a reviewer spends getting the patch ready to merge.
If Carbon finishes in one pass and Argon needs retries, the per-token rate tells only part of the story. Repeat the comparison with December rates before moving a job that runs every day. This is a proposed evaluation. We have no head-to-head measurements for that workload.
Abdessalam Alaoui
I'm Abdessalam Alaoui, founder of Musthave.AI. I write practical guides to AI models, tools, generative media.
