Gemini 4 Hydrogen

Runs MCP tools

The efficiency model of the Gemini 4 family, listed on OpenRouter from 12 November 2026. It is built for high-volume agent work, coding and model routing, at lower cost and latency than Carbon and Argon.

Vendor
Google DeepMind
Released
12 November 2026
Context
1.05M tokens
Input
Text, image, video, audio, file

Paste a server URL, pick a model, and watch it call your tools in a real conversation. No install.

No server of your own? Leave it blank and use one of the public mock servers.

In your browser

MCP Agent Studio

Server has a public URL? Chat with Gemini 4 Hydrogen on it, watch every tool call, and compare models side by side.

Open MCP Agent Studio →

On your computer

MCP Playground Desktop

Server runs on localhost or over stdio? The browser can’t reach it. The desktop app can, with Gemini 4 Hydrogen in Agent Studio. Download free

MCP Evals

Can Gemini 4 Hydrogen actually complete tasks with your server?

One chat shows the tool calls work. An eval writes real tasks from your tool schemas, has Gemini 4 Hydrogen drive each one, and reports where it got the wrong answer even though every call succeeded. Pick Gemini 4 Hydrogen as the driver.

What Gemini 4 Hydrogen is

OpenRouter describes Gemini 4 Hydrogen as the high-efficiency workhorse of the Gemini 4 family, built for high-throughput agentic workflows, coding and multi-model routing. It has the same 1M-token window and text, image, video, audio and file input as Carbon and Argon, with tools, structured outputs and low to high reasoning effort. It is the cheapest Gemini 4 model on OpenRouter. Before release it circulated as a codename for a fast, Flash-tier Gemini 4, but we found no Google announcement or model card for it yet, so there are no vendor benchmarks to quote.

How it reaches MCP

  • MCP Playground’s Agent Studio: paste your server URL above and run Gemini 4 Hydrogen against it in the browser, with no Google key
  • OpenRouter’s API with any MCP client that turns MCP tools into function calls

What to watch for

  • There are no published benchmarks for Hydrogen yet. Judge it on your own server: run your eval suite on Hydrogen and on Carbon or Argon, and keep Hydrogen only where the results match.
  • Fast, lower-cost models tend to give up sooner on long tool chains. Test the longest chain your server needs, not only single calls.
  • A model with no launch post can change behaviour or name. Pin google/gemini-4-hydrogen and re-run your evals when Google publishes its model card.

Run Hydrogen and Carbon on the same four-step task against the complex-schema mock: create_user_profile, process_order, analyze_data, then configure_workflow. Compare where each one first sends a wrong argument, and how long each run takes. All the mock servers →

Available here

ModelModel ID
Gemini 4 Hydrogengoogle/gemini-4-hydrogen

Go deeper

Frequently asked questions

Does Gemini 4 Hydrogen support tool calling?
Yes. OpenRouter lists tools, tool_choice and structured_outputs. Any MCP client that turns MCP tools into function calls can drive it, and you can run it against your own server here without a Google key.
Is Gemini 4 Hydrogen the same as Gemini 4 Flash?
Hydrogen circulated as a codename for a fast, Flash-tier Gemini 4 before release, and OpenRouter now lists it under the Hydrogen name. We found no Google announcement confirming how it relates to the Flash line, so treat the two names as unconfirmed.
Gemini 4 Hydrogen or Gemini 4 Carbon?
Hydrogen is the cheaper, faster model for high-volume tool calls and routing. Carbon is Google’s strongest model for long, hard runs. Try Hydrogen first for simple, repeated tool use, and switch to Carbon if your evals show it missing steps.

Sources