# Gemini 4 Hydrogen

> Gemini 4 Hydrogen, released by Google on November 12, 2026, targets everyday coding and enterprise agent workflows. Google emphasizes speed and efficiency, alongside multimodal analysis and tool use. Where Carbon targets deep work on complex tasks, Hydrogen is a candidate for repeated development and agent routines. Compare your own prompts, tool steps and output lengths before concluding that a workload will finish faster or cost less.

Canonical: https://routemux.com/google/gemini-4-hydrogen
Provider: Google

## Key facts

| Field | Value |
| --- | --- |
| Model ID | google/gemini-4-hydrogen |
| Official model ID | gemini-4-hydrogen |
| Context window | 1M |
| Max output | 1M |
| Input modalities | text, image, audio |
| Output modalities | text |
| Capabilities | Vision, Reasoning, Tool Calling, Json Mode, Coding, Audio Input |
| Upstream release | 2026-11-12 |
| Availability | available |

## Pricing

- RouteMux input: $0.75 per 1M tokens
- RouteMux output: $3.75 per 1M tokens
- Official input: $1.5 per 1M tokens
- Official output: $7.5 per 1M tokens
- Cache read: $0.0375 per 1M tokens
- Saving vs official: 50%
- Official price source: https://cloud.google.com/model-garden/pricing, verified 2026-11-14

Billing is prepaid wallet, charged per successful request only.

## API access

Base URL: https://api.routemux.com

Supported protocols:

- openai_chat
- openai_responses
- anthropic_messages

Aliases that also route here: models/gemini-4-hydrogen

## What is Gemini 4 Hydrogen

Gemini 4 Hydrogen, released by Google on November 12, 2026, targets everyday coding and enterprise agent workflows. Google emphasizes speed and efficiency, alongside multimodal analysis and tool use. Where Carbon targets deep work on complex tasks, Hydrogen is a candidate for repeated development and agent routines. Compare your own prompts, tool steps and output lengths before concluding that a workload will finish faster or cost less.

Hydrogen defaults to high thinking and offers low, medium, high and xhigh levels. Thinking consumes output tokens, so the length of the visible answer is not the full usage. Input, thinking, function calls and the final answer share a 1,048,576-token window. Hydrogen supports up to 65,536 output tokens in one segment with or without streaming, unlike Carbon's 32,768-token non-streaming cap. Google's million-token output figure still requires continuation between segments; the compatibility APIs do not promise automatic continuation.

RouteMux exposes OpenAI Chat Completions and Anthropic Messages with streaming, ordinary function tools and Base64 images. Office documents, audio, video and hosted tools listed for Google's native APIs do not automatically carry over to these endpoints. Custom temperature, top_p and top_k values are ignored. Cache billing follows actual implicit-cache usage returned by the model; an Anthropic cache_control hint does not establish explicit cache writes.

RouteMux input costs $0.75 per million tokens and output costs $3.75 per million tokens.

## Key capabilities

- **Everyday coding and agents** — Google positions Hydrogen around speed and efficiency for everyday coding, enterprise agents, multimodal analysis and tool use.
- **High by default, plus xhigh** — The model offers low, medium, high and xhigh. Thinking and the answer share an output budget; consult the published parameter matrix for values accepted by each compatibility endpoint.
- **65,536 tokens in either mode** — Both streaming and non-streaming requests have a 65,536-token segment cap. Million-token output requires continuation across segments.
- **Images alongside tool history** — Compatibility endpoints accept Base64 images and ordinary function tools, including images inside tool results. Check endpoint support separately for office files and Google-hosted tools.

## Use cases

- **Everyday development** — A use named by Google: integrate code analysis, changes and checks into repeated development tasks.
- **Enterprise agent workflows** — Another stated use: progress through business tasks with several tool calls and continue from the returned results.
- **Multimodal analysis** — Google includes multimodal analysis in Hydrogen’s positioning. Base64 image understanding and images returned by tools are supported starting points on these compatibility endpoints.

## Other variants in this family

- Gemini 4 Argon (`google/gemini-4-argon`)
- Gemini 4 Carbon (`google/gemini-4-carbon`)

## Sources

- [Google: Gemini 4 Hydrogen model documentation](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/4-hydrogen)
- [Google: Get started with Gemini 4](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/start/get-started-with-gemini-4)