Gemini 4 Carbon rolls out: what ecommerce teams can test
Google opens paying API access and starts an AI Ultra rollout. A practical way to evaluate Carbon without confusing model access with a commerce launch.
Gemini 4 Carbon is moving beyond its initial restricted release. Google’s 10 November announcement opens API access to paying developers and enterprises, with a rollout to Google AI Ultra subscribers in the Gemini app, Antigravity and AI Mode in Search.
For ecommerce teams, there are two separate opportunities to assess: using the model for internal work and observing answers generated in an eligible AI Mode session. Access to one surface should not be taken as evidence of access to every other surface.
Availability is the change to act on
Google positions Carbon for difficult, extended tasks and says Argon remains suitable for most software engineering and knowledge work. It is also previewing Antigravity’s /swarm feature, which coordinates agents on a project.
The announced introductory API price is $2 per million input tokens and $10 per million output tokens. That is token pricing, not a fixed price for completing a store task: Google says Carbon uses more tokens than Argon for deeper reasoning and verification.
Start with a bounded store task
Our recommendation is to compare models on one repeatable job before changing a production workflow. A catalogue-quality report is a useful starting point: supply the same product records, define the fields that need checking and keep a human-approved answer set.
Measure errors and review time as well as the bill. Require the output to identify missing information rather than invent product specifications. Keep any proposed changes in a review file until a person has checked them.
For AI Mode, record the account, question and date when reviewing a shopping answer. A different answer is an observation, not proof that a model change caused a ranking gain or loss. This rollout should not be treated as a new Shopify checkout integration or a promise of more product referrals.
The practical question is whether Carbon improves a specific piece of work enough to justify its cost and supervision. The launch announcement alone cannot answer that for your store.
For a broader evaluation framework, see our guide to comparing AI models for Shopify workflows.
Gemini 4 Carbon is now available.
November 10, 2026
Source: Google, 10 November 2026. If the embedded announcement does not load, read the original post.
16 November follow-up: Pro access and the missing model label
Google Search engineering VP Rajan Patel confirmed two changes in replies dated 15 November: Carbon access appearing for Google AI Pro subscribers in AI Mode was not a bug, and removing the selected-model indicator was intentional. He said Google was considering how to restore an indicator. Search Engine Roundtable documented the exchange on 16 November.
For teams checking product answers, record the account tier and test date. The absence of a model name in a screenshot is not proof of which model answered, and broader access does not establish that a particular product will appear or rank differently.
Rajan Patel on the model indicator — 15 November 2026
More from AI
See all →
