Google has expanded the Gemini 4 lineup by introducing the new Hydrogen model, which the company positions as a faster and more cost-effective replacement for the previous 3.8 Flash version. Unlike earlier releases, access to which was initially restricted to a narrow circle of cybersecurity specialists and holders of the most expensive subscriptions, the new model opens to users immediately after the announcement. According to the company, Hydrogen significantly outperforms 3.8 Flash across all key parameters and will serve as the basis for a series of changes in Google Search.
Four Levels of Thinking Depth and the xhigh Mode
The main structural difference between Hydrogen and its predecessor is the support for four levels of "thinking" depth: low, medium, high, and the new extra-high (xhigh) mode. Even in the high mode, the model operates 2.5 times faster than the 3.8 Flash version, while the number of fabricated statements—so-called hallucinations—has been reduced by more than five times. This set of levels allows developers and users to flexibly balance response speed, reasoning quality, and processing cost.
Agentic Coding and Outperforming the Flagship Argon
For developers, Hydrogen becomes the main tool for agentic coding: the model tripled its predecessor's results in the Terminal-Bench test and is capable of quickly generating basic code and running tests without excessive token consumption. In the extra-high thinking mode, Hydrogen even outperformed the older Argon model in the popular Artificial Analysis Intelligence Index ranking, scoring 54 points against the flagship's 50.
Largest Update to Search AI Tools
One of the major changes is the integration of Hydrogen into Google Search, which the company calls the largest update to its search AI tools in history. The model handles complex multi-part queries better, adds more visual elements in AI Overviews and AI Mode, and is tuned to provide more source links. In Google Search itself, the updated model will be available to all users without exception.
Access and Pricing: Who Can Use Hydrogen Now
In the Gemini app and AI Studio, Hydrogen is currently only available to paid AI Pro and Ultra subscribers, while free version and AI Plus users remain on the older 3.5 Flash-Lite model. For developers via the API, processing costs remain at the 3.8 Flash level, standing at $1.50 per million input tokens and $7.50 per million output tokens. By comparison, prices for the higher-end Carbon and Argon models will increase next month from $2/$10 to $4/$20, making Hydrogen an economically attractive option for mass-market scenarios.
