All posts
AI

Google Releases Gemini 4 Hydrogen at $1.50 per Million Input Tokens, Replacing 3.8 Flash

Google: Gemini 4 Hydrogen announcement

Google released Gemini 4 Hydrogen on 12 November 2026. Koray Kavukcuoglu, who announced it, calls it Google's daily driver for agentic workflows. It sits below the heavier Gemini 4 Carbon and Gemini 4 Argon models, which handle long, deep reasoning, and Google positions it as the upgrade for teams on Gemini 3.8 Flash.

The pricing is the first thing to note. Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, flat across the full 1 million token context window. Output can also run to 1 million tokens.

There are four thinking levels: low, medium, high (the default) and a new xhigh tier. Google says Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default, against 50 for Argon, and 54 at xhigh. On Terminal-Bench 4.1 it scores 57 percent, against 18.2 percent for 3.8 Flash, and Google says it finishes tasks about 2.5 times faster than the older model.

It is rolling out widely. Developers get it in the Gemini API, Google AI Studio and Android Studio. Businesses get it through Gemini for business and in Google Sheets for eligible plans. Consumers get it in the Gemini app and in Google Search for all users globally, with free app users reaching it through automatic model selection.

For a design and web studio, the most relevant claims are about front-end work. Google says Hydrogen produces more complete, functional web layouts in fewer prompts, and shows examples including a one-shot e-commerce storefront with WebGL2 shaders and an interactive 3D globe. Google also reports that Hydrogen cuts overreach in agentic tasks by up to 81 percent, meaning fewer cases of the model doing more than it was asked.

We treat demo galleries with caution. A storefront generated in one prompt is a sketch, not a site, and the hard part of web work is still content, accessibility, performance and maintenance. Where a fast, cheap model does help is the early stage: quick layout options to react to, throwaway prototypes for a client call, and small refactors.

The flat price across the whole context window is the more meaningful change for production use. Long-context jobs, such as reading an entire brand guideline PDF and a site's codebase together, often became expensive at the top end of a context window. If Google's numbers hold, Hydrogen makes those jobs cheap enough to run routinely.

As always, we will test it against the models we already use on real client tasks before switching anything.