Gemini 4 Neon now available on AI Gateway
- Source
- vercel.com
- Author
- Jerilyn Zheng
- Date

Context
Developers can now call Gemini 4 Neon through Vercel's AI Gateway by using the model ID `google/gemini-4-neon`, according to Vercel's changelog. AI Gateway is a hosted API that sits between an application and many model providers. The application authenticates once with a gateway key, and the gateway forwards each request, logs it, tracks its cost and can enforce budgets. Vercel's documentation says the gateway adds no markup to provider token prices, and the model page lists Neon at $3 per million input tokens and $15 per million output tokens, the same list price Google announced.
What the gateway adds is routing. The model page lists two providers for Neon, Google's own API and Vertex AI, Google Cloud's model platform, and the gateway can send a request to a healthy provider and fall back to another provider or another model when one fails. A team that already reaches other labs' models through the gateway can try Neon by changing one string in its code. The changelog shows that with the AI SDK, Vercel's TypeScript library, and also gives setups for Claude Code and for apps that speak OpenAI's Chat Completions format.
The listing also reaches coding agents that Google does not make. Vercel's coding agents guide, which the changelog links for Codex, Cursor and other tools, points those agents at the gateway with a single setup command, after which they can call Neon by the same model ID; Codex and Claude Code, for example, list the gateway's models in their model pickers. The changelog says the model takes text, images, video and documents, supports , structured outputs and with Google Search and Google Maps, and exposes the same four reasoning levels, low, medium, high and xhigh.
- tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
- grounding — Tying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
- multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
Checking sign-in…
Loading comments…






