Gemini 4 Carbon now available on AI Gateway

I am sizing each paragraph to that material so the story stays inside 800–1,000 words.Google's Gemini 4 Carbon is now available on Vercel AI Gateway, and Vercel Blog's report lists the model at 50% off through November 30, 2026. Google built Carbon for complex software engineering, sustained agentic work, and cybersecurity defense. The Gateway listing gives the model a 1M-token context window for large codebases and long inputs, and it publishes the caller name google/gemini-4-carbon.

I am sizing each paragraph to that material so the story stays inside 800–1,000 words.Google's Gemini 4 Carbon is now available on Vercel AI Gateway, and Vercel Blog's report lists the model at 50% off through November 30, 2026. Google built Carbon for complex software engineering, sustained agentic work, and cybersecurity defense. The Gateway listing gives the model a 1M-token context window for large codebases and long inputs, and it publishes the caller name google/gemini-4-carbon.

This piece is for developers who call AI Gateway from the AI SDK, Claude Code, Codex, or Cursor. It covers the 50% discount through November 30, 2026, the 1M-token context window, and the engineering, cybersecurity, and multi-step behavior described for Carbon. It also covers the setup command and AI Gateway's usage and cost tracking, retries, failover, Zero Data Retention, budgets for API keys, routing rules, no markup, and no platform fee on Bring Your Own Key requests.

01Half-price access through November 30

Vercel is offering Google's Gemini 4 Carbon on AI Gateway at 50% off through November 30, 2026. The changelog sets the model name to google/gemini-4-carbon and ties that 50% price to the November date. Developers who send google/gemini-4-carbon while this window is open get the 50% off rate the listing states for the model.

Google built Gemini 4 Carbon for complex software engineering, sustained agentic work, and cybersecurity defense. Vercel is listing that Google model on AI Gateway, and those three jobs are the scope written next to the discount. The November 30, 2026 end date applies to google/gemini-4-carbon for an AI SDK call and for Claude Code, Codex, Cursor, or another agent.

02One million tokens for codebases

Gemini 4 Carbon has a 1M-token context window for working with large codebases and long inputs. The listing attaches that 1M-token figure to those two cases, a large codebase and a long input, and that is the context size it gives this model on AI Gateway. On engineering tasks, Carbon carries demanding work through multiple stages.

Those stages come with improvements in coding, full-stack web development, and scientific reasoning. Sustained agentic work is the second of the three jobs Google built Carbon for, named alongside complex software engineering and cybersecurity defense. A request that uses the 1M-token window for a large codebase or a long input is addressed to google/gemini-4-carbon on AI Gateway.

03Coding gains and vulnerability patches

Coding, full-stack web development, and scientific reasoning are the engineering improvements the listing names for Gemini 4 Carbon. Vercel places all three on demanding work that the model carries through multiple stages. A large codebase or a long input on that staged work uses the 1M-token context window published for the model.

Carbon's cybersecurity capabilities span finding vulnerabilities, validating their risk, and developing patches. Cybersecurity defense is the third job Google built into the model, beside complex software engineering and sustained agentic work. Vercel lists that cybersecurity work on AI Gateway under Gemini 4 Carbon, at the model name google/gemini-4-carbon.

Google reports more reliable execution of multi-step plans, fewer unsupported claims, and better collaboration when requirements are ambiguous. The listing says this helps agents follow through while respecting project constraints. The figures on the same listing are a 50% discount through November 30, 2026, provider pricing with no markup, no platform fee on inference including Bring Your Own Key (BYOK) requests, and a 1M-token context window.

04Setup for Claude Code and Cursor

Developers pass google/gemini-4-carbon as the model name on AI Gateway. In the AI SDK sample, streamText is imported from the ai package, the model field is set to that string, and the prompt asks for a TypeScript helper that safely parses JSON and returns a fallback. A loop then reads each chunk from result.textStream and writes it with process.stdout.write.

```javascript
import { streamText } from 'ai';

Claude Code, Codex, Cursor, and other agents follow a Vercel CLI path. Install the latest Vercel CLI, run npx vercel ai-gateway setup, and select google/gemini-4-carbon in the agent. Vercel also lists Gemini 4 Carbon in the model playground, names the coding agents guide for those agent steps, and labels Claude Code and Chat Completions beside the AI SDK sample.

npx vercel ai-gateway setup

05Zero markup and key budgets

AI Gateway is a unified API for calling models, tracking usage and cost, and configuring retries, failover, and performance optimizations for higher-than-provider uptime. The product includes built-in custom reporting, Zero Data Retention support, budgets for API keys, and routing rules. Those controls sit on the same Gateway that lists Gemini 4 Carbon as google/gemini-4-carbon.

AI Gateway reflects provider pricing with no markup on the models it serves, including Gemini 4 Carbon. It charges no platform fee on inference, and that rule includes Bring Your Own Key (BYOK) requests. Vercel says every language model available on AI Gateway can be viewed in the catalog, with google/gemini-4-carbon among them.

November 30, 2026 is the last day Vercel attaches to the 50% off rate for Gemini 4 Carbon on AI Gateway. After npx vercel ai-gateway setup, Claude Code, Codex, Cursor, and other agents select google/gemini-4-carbon. The context window on that offer is 1M tokens, the Gateway adds no markup and no platform fee, and that rule includes Bring Your Own Key requests.