The model is available at 50% off until November 30, 2026. For those building sites with AI coding tools and already routing requests through Vercel, it offers a new model to try at a lower price for about three weeks.
This applies to people who call models through Vercel AI Gateway or configure a coding agent using npx vercel ai-gateway setup. The changelog does not mention v0, so it does not inform v0 users in the builder about a change.
What is new
- Gemini 4 Carbon from Google is now listed on AI Gateway under the name google/gemini-4-carbon.
- It is priced at 50% off through November 30, 2026.
- It has a 1-million-token context window, aimed at large codebases and long inputs.
- Coding agents can use it after you install the latest Vercel CLI and run npx vercel ai-gateway setup.
Taken from the official update: vercel.com [1]
Gemini 4 Carbon is aimed at engineering and agent work
Vercel says the model is designed for complex software engineering, extended agentic work and cybersecurity defense. In plain terms, it is aimed at lengthy coding tasks that require an AI agent to work through many steps, rather than quick, one-line answers.
Fact checked on 11 Nov 2026 against vercel.com [1]
If you work on a large project, remember the 1-million-token context window. It lets you provide a large codebase or lengthy documents in one request. The model page also lists a maximum output of 1,000,000 tokens, with reasoning tokens counting toward that limit, so extended reasoning leaves less room for the answer.
Fact checked on 11 Nov 2026 against vercel.com [1], vercel.com [2]
The model page lists text, images and PDFs as inputs. Images and PDFs count as input tokens, which matters when estimating the cost of a request that includes design screenshots or a PDF brief.
Fact checked on 11 Nov 2026 against vercel.com [2]
The 50% discount ends on November 30, 2026
The model page lists prices starting at $2 per million input tokens and $10 per million output tokens, with a 50% off label. Vercel's changelog says the discount lasts through November 30, 2026. Since the pages do not list prices after that date, treat the current figures as temporary rather than the long term cost.
Fact checked on 11 Nov 2026 against vercel.com [2], vercel.com [1]
The word "from" is significant because there is more than one rate. For the Google provider, the AI Gateway endpoint listing gives standard rates of $2 and $10, rates of $3.60 for input and $18 for output when a request is served at priority, which offers higher availability and faster processing, and rates of $1 for input and $5 for output when served at flex, which costs less and may have higher latency. For Vertex AI, the listing gives base rates of $2 and $10, while regional inference costs $2.20 and $11.
Fact checked on 11 Nov 2026 against ai-gateway.vercel.sh [4]
AI Gateway matches provider pricing without a markup and charges no platform fee for inference, even when requests use your own provider key. That means the price shown is the model provider’s rate, with no added Vercel charge.
Fact checked on 11 Nov 2026 against vercel.com [1]
Plans: the gateway is on every plan, but the free tier is limited
Vercel’s documentation says AI Gateway is available on every plan. According to the pricing page, the free tier provides a $5 monthly credit for eligible models only, while the paid tier includes access to all available models.
Fact checked on 11 Nov 2026 against vercel.com [3]
Neither page confirms whether Gemini 4 Carbon is included among the eligible free tier models. If you use the free tier, check the model’s eligibility before planning work around it.
Fact checked on 11 Nov 2026 against vercel.com [3]
What the official notes do not say
- The changelog and model page do not say whether Gemini 4 Carbon is eligible for the free AI Gateway tier.
- They do not say which model or feature Gemini 4 Carbon replaces.
- They do not give a rate limit or quota that is specific to this model.
- They do not give a price for after November 30, 2026.
- The changelog does not mention v0, so it does not say whether v0 itself offers this model.
How it helps: a freelancer who builds client sites with an AI coding agent such as Claude Code, Codex or Cursor
The task. Try Gemini 4 Carbon on a larger site project before the discount ends
- Install the latest Vercel CLI.
- Run npx vercel ai-gateway setup in your terminal.
- Open your coding agent and select google/gemini-4-carbon as the model.
- Give it a task that needs a lot of context, such as a whole theme or a long brief. Images and PDFs in the brief count as input tokens, and reasoning counts toward the output limit, so a large brief costs more at the from prices.
The result. Your agent sends requests to Gemini 4 Carbon through AI Gateway at the discounted rate, with no Vercel markup, through November 30, 2026. On the free tier, the model may not be available unless it is an eligible model.
Fact checked on 11 Nov 2026 against vercel.com [1], vercel.com [2], vercel.com [3]
Other ways to use it
In the AI SDK, set the model to google/gemini-4-carbon in a streamText call. Vercel’s example asks the model to write a TypeScript helper for safely parsing JSON. The changelog also links to a model playground for trying it first.
Fact checked on 11 Nov 2026 against vercel.com [1]
Sources
- https://vercel.com/changelog/gemini-4-carbon-now-available-on-ai-gateway (official update)
- https://vercel.com/ai-gateway/models/gemini-4-carbon
- https://vercel.com/docs/ai-gateway/pricing
- https://ai-gateway.vercel.sh/v1/models/google/gemini-4-carbon/endpoints
- https://vercel.com/ai-gateway/models/gemini-4-argon
Facts in this article were checked against these sources on 11 Nov 2026. Product details change, so check the official page before you rely on them.