Google made Gemini 4 Hydrogen (model ID gemini-4-hydrogen) generally available on November 12, 2026. It is the third Gemini 4 model after Argon (October 13) and Carbon (November 10), and the first Gemini 4 model you can use on the free tier.
The short version: paid-tier pricing is $1.50 per 1M input tokens and $7.50 per 1M output tokens, below Carbon and Argon ($2.00 / $10.00 through November 30, then $4.00 / $20.00 from December 1). Google calls Hydrogen the recommended upgrade from Gemini 3.8 Flash and earlier Gemini 3 and 2.5 models. Everything below is taken from the official Gemini API changelog, pricing page and the "What's new in Gemini 4 Hydrogen" guide. Checked: November 13, 2026.
Gemini 4 Hydrogen at a glance
Item | Details |
|---|---|
Model ID |
|
Release | GA on November 12, 2026 (no preview). No shutdown date announced |
Price (Standard, paid tier) | $1.50 input / $7.50 output per 1M tokens (thinking tokens billed as output). Same rate across the full 1M context |
Caching | Implicit caching only: $0.075 per 1M cached input tokens (95% off). Explicit caching ( |
Batch | $0.75 input / $3.75 output |
Free tier | Yes, in Google AI Studio and the Gemini API (rate-limited; free-tier inputs may be used to improve Google's products) |
Context | 1,048,576 tokens shared by input, thinking and output |
Max output | 65,536 tokens per synchronous request; up to 1M with long decode (background execution) |
Thinking levels |
|
Not supported | Image generation, audio generation, Live API |
Hydrogen vs Carbon vs Argon
Model | Positioning (Google) | Price (in / out) | Default thinking | Free tier |
|---|---|---|---|---|
Gemini 4 Carbon | Strongest reasoning for hard architecture, debugging and security audits | $2.00 / $10.00 until Nov 30, then $4.00 / $20.00 | medium (max high) | No |
Gemini 4 Argon | Most advanced long-horizon tasks with fewer tokens than Carbon | Same as Carbon | medium (max high) | No |
Gemini 4 Hydrogen | Fast, efficient default for everyday coding, agents and document work | $1.50 / $7.50 | high (max xhigh) | Yes |
Google also suggests pairing them: let Carbon plan or resolve hard roadblocks, then hand code edits, parallel subagent tasks and test runs to Hydrogen. With the Interactions API you can switch models between turns of one session using previous_interaction_id.
When to use xhigh
Hydrogen adds a fourth thinking level, xhigh, for hard multi-step coding and agentic problems where high falls short. Google warns that it raises latency and token use and can overthink simple prompts. The official advice is to evaluate all four levels and pick the lowest one that meets your quality bar. Because thinking tokens are billed at the output rate, the default high setting alone can make real costs higher than a price-sheet estimate.
Migrating from Gemini 3.8 Flash
- Change the model string to
gemini-4-hydrogen. - Remove hardcoded low
max_output_tokenscaps (for example 8,192) or raise them to 65,536. - Set
thinking_levelexplicitly; replaceminimalwithlow. - Drop
cached_content; implicit caching applies automatically to prompts of 4,096+ tokens. - Remove
temperature,top_pandtop_k; the server ignores them.
Cost note: Gemini 3.8 Flash costs $0.75 / $3.75 until December 31, 2026 and rises to $1.50 / $7.50 on January 1, 2027, the same as Hydrogen. Testing in December and switching on January 1 avoids paying more than necessary.
FAQ
How much does Gemini 4 Hydrogen cost?
On the paid tier, $1.50 per 1M input tokens and $7.50 per 1M output tokens, with thinking tokens billed as output. Batch is $0.75 and $3.75, and implicitly cached input is $0.075.
Is Gemini 4 Hydrogen free?
Yes, it is available on the free tier of Google AI Studio and the Gemini API with rate limits. It is the only Gemini 4 model with a free tier; free-tier inputs may be used to improve Google's products.
What is the xhigh thinking level?
A new fourth thinking level in Hydrogen that spends the most tokens for the deepest reasoning. The default is high, and Google recommends xhigh only for hard multi-step problems where high falls short.