Gemini 4 Hydrogen is available on BetterToken
Use Gemini 4 Hydrogen for everyday coding and multi-step tasks. Find the Model ID, API setup, official token limits, thinking levels, and caching details.
Updated: November 13, 2026
Gemini 4 Hydrogen is available on BetterToken. Use your existing BetterToken API Key with Model ID gemini-4-hydrogen.
When to use Hydrogen
Hydrogen focuses on speed and efficiency for everyday coding and multi-step tasks. Consider it for code edits, UI prototypes, and repeated development tasks.
Google sets the default thinking level to high and adds xhigh for harder problems. Higher thinking levels can increase latency and token use.
Official specifications
| Property | Google specification |
|---|---|
| Model ID | gemini-4-hydrogen |
| Input | Text, images, video, audio, and PDF |
| Output | Text |
| Unified context | Input, thinking, and output share 1,048,576 tokens |
| Default synchronous output limit | 65,536 tokens |
| Thinking levels | low, medium, high (default), xhigh |
| Caching | Implicit caching only; explicit cache creation is not supported |
BetterToken setup
Hydrogen supports non-streaming and streaming text through Chat Completions and native Gemini. The native interface also supports structured JSON and function calling.
Use your existing BetterToken API Key, or register and create one.
| Protocol | Base URL | Authentication header |
|---|---|---|
| Native Gemini / Google GenAI SDK | https://bettertoken.ai | x-goog-api-key: YOUR_BETTERTOKEN_API_KEY |
| OpenAI-compatible Chat Completions | https://bettertoken.ai/v1 | Authorization: Bearer YOUR_BETTERTOKEN_API_KEY |
Native requests use POST /v1beta/models/gemini-4-hydrogen:generateContent. Do not append /v1beta to the SDK Base URL. Chat Completions uses POST /v1/chat/completions with model set to gemini-4-hydrogen.
For setup and text request examples, see the Gemini API guide.
Pricing and caching
Check the model plaza for current input, output, and cache-read prices. Thinking tokens count as output. Implicit caching works automatically, without creating cache objects.

