Google has launched Gemini 4 Hydrogen, the least expensive member of the Gemini 4 family. Its price is not far from that of its two more powerful models, Carbon and Argon, but the entire pricing grid remains, according to the analysis, difficult to read.
A grid that changes on December 1st
Gemini 4 Carbon and Gemini 4 Argon are charged at the same rate: until November 30, 2026, $2.00 per million input tokens, $0.10 for cached inputs, and $10.00 for output. Starting December 1, 2026, these prices double to $4.00, $0.20, and $20.00. A token refers to a unit of text processed by the model, and caching allows for the reuse of previously sent content at a lower cost.
This raises the question: why offer two models at the same price? Google states that Argon ‘delivers Gemini 4 intelligence with faster responses and lower token consumption per turn than Gemini 4 Carbon.’ It can therefore cost the same per unit while using fewer tokens to accomplish the same task.
Hydrogen, an entry-level model
Gemini 4 Hydrogen is priced at $1.50 per million input tokens, $0.075 for cached input, and $7.50 for output. Compared to Carbon and Argon, the reduction is currently only 25%. It should become more noticeable after December 1, 2026, when the other two models increase in price.
Hydrogen also remains more expensive than Claude Haiku 5.5 and GPT-6 Luna, both priced at $0.10 for input and $0.50 for output. Hydrogen’s input prices are thus fifteen times higher than those of these two models.
To compare response quality, the analysis asked Hydrogen to generate images of pelicans. A full grid presents the results of the three Gemini 4 family models at their three supported effort levels. According to the analysis, the quality gap between the pelicans produced by Hydrogen and those of the other two models is ‘quite clear,’ even though their prices are close.
The role assigned to it by Google
Google describes Hydrogen’s place in its lineup by distinguishing use cases. In software development, teams rely on cutting-edge models like Carbon and Argon for architecture choices and solving complex bugs. They also need, according to Google, a fast and rigorous execution model for writing code, running tests, and checking changes. Google specifies that Hydrogen ‘is designed to stand alone as a model for daily work, or to be paired with Carbon and Argon as a highly efficient execution engine in multi-model workflows.’ Hydrogen is also the new model for the free offering. It will power the AI features in Google Search.
