Create each key on its own page, paste it below, then copy the finished lines into your project's .env file. What you paste stays in this tab. It is never sent to Wirestack or saved.
On November 12, 2026 Google made Gemini 4 Hydrogen (gemini-4-hydrogen) generally available in the Gemini API, on both the GenerateContent and Interactions APIs.
Free tier and paid tier. The free tier costs nothing and is rate-limited.
Paid price, per 1M tokens: $1.50 input, $7.50 output (thinking tokens count as output), $0.075 cached input through implicit caching.
Context: 1,048,576 tokens shared by input, thinking and output. A normal response stops at 65,536 tokens. With background: true and a higher max_output_tokens it can run longer.
Thinking:low, medium, high (the default) and a new xhigh.
The Antigravity agent in Gemini Managed Agents now uses Hydrogen by default. You can change the model with agent_config.
What can break when you switch from Gemini 3.8 Flash or earlier:
The default thinking level is high, not medium, so the same request can think longer and cost more. minimal is rejected with an error.
temperature, top_p and top_k are ignored. Remove them.
Explicit caching is not supported (cachedContents, caches.create, cached_content). Remove those calls. Implicit caching applies to prompts of 4,096 tokens or more.
Asking for tool calls as raw-text JSON with no tools declared can return MALFORMED_FUNCTION_CALL. Declare tools or use structured outputs.
Coming from Gemini 4 Carbon or Argon, the request shape is the same. Those two default to medium and stop at high.
Prerequisites
A Gemini API key in GEMINI_API_KEY
The Google Gen AI SDK (google-genai for Python)
Steps
Change the model ID to gemini-4-hydrogen.
Remove temperature, top_p, top_k and any explicit cache calls.
If you sent the minimal thinking level, change it to low. Set a level yourself if you do not want the high default.
Remove low max_output_tokens caps such as 8,192, or raise them to 65,536.
Use background mode for outputs longer than 65,536 tokens.
Google's Python quickstart (Interactions API):
Python
from google import genaiclient = genai.Client()interaction = client.interactions.create( model="gemini-4-hydrogen", input="Build an interactive single-file HTML token bucket rate-limiter simulator with live sliders for burst size and request traffic.")print(interaction.output_text)
Test it
Run the quickstart with your key. Then send one of your own prompts and compare the response time and token count with the model you use today, since the default thinking level is higher.
Not confirmed
The page does not state a minimum SDK version.
The free tier's rate limits were not read on the rate limits page.
Wirestack