November 12, 2026
Release llm-gemini 0.37 — LLM plugin to access Google’s Gemini family of models
Google released Gemini 4 Hydrogen today — the lowest-cost member of the Gemini 4 family.
It’s actually not that much cheaper than its larger siblings, Carbon and Argon. I find the Gemini 4 pricing strategy very confusing. Here is how it breaks down:
| Model | Date | Input ($/M tokens) | Cached Input | Output |
|---|---|---|---|---|
| Gemini 4 Carbon | Until Nov 30, 2026 | $2.00 | $0.10 | $10.00 |
| From Dec 1, 2026 | $4.00 | $0.20 | $20.00 | |
| Gemini 4 Argon | Until Nov 30, 2026 | $2.00 | $0.10 | $10.00 |
| From Dec 1, 2026 | $4.00 | $0.20 | $20.00 | |
| Gemini 4 Hydrogen | $1.50 | $0.075 | $7.50 |
Carbon and Argon are the exact same price, and both are scheduled to double in price on December 1st. Why have two different models at the same price? Google says that compared to Gemini 4 Carbon, Argon “delivers Gemini 4-class intelligence with faster responses and fewer tokens per turn” — so while the price per token is the same, it will consume fewer tokens per task.
Hydrogen is ostensibly the cheap model, yet right now it’s only 25% cheaper than its more capable siblings. Hydrogen’s pricing will make a lot more sense after December 1st, when it will represent a 75% discount (one-quarter the price). Even then, it remains significantly more expensive than Claude Haiku 5.5 and GPT-6 Luna — both of which are priced at $0.10/$0.50 for input/output, making Hydrogen 15 times more expensive.
I had Hydrogen generate some pelicans. Here is a full comparison grid showing the pelicans generated by all three members of the Gemini 4 family across the three supported effort levels:

Given how close they are in price, the drop-off in pelican quality for Hydrogen is quite dramatic.
Here is how Google describes Hydrogen’s intended role in their lineup:
In real-world software development, while teams rely on frontier models like Carbon and Argon for deep architectural judgment when designing systems or debugging gnarly bugs, they still need a fast, disciplined execution model to write the code, run the tests, and verify the changes. Hydrogen works both as a standalone model for everyday work and as a high-velocity execution engine paired with Carbon and Argon in multi-model workflows.
Most notably, Hydrogen is the new free tier model and will power the AI features in Google Search.