Google launches Gemini 4 Hydrogen: a fast, cheap model it says beats flagship Argon
Gemini 4 Hydrogen costs $1.50 per million input tokens, runs 2.5 times faster than 3.8 Flash and is rolling out in Google Search for all users globally.


Google on Thursday evening launched Gemini 4 Hydrogen, a fast and cheaper model in the Gemini 4 family. The model is rolling out in Google Search for all users globally, in the Gemini app and to developers through the Gemini API. Google says it beats its own flagship Gemini 4 Argon on an independent intelligence index and finishes tasks 2.5 times faster than its predecessor, Gemini 3.8 Flash.
Google published the announcement on its blog on Thursday 12 November at 19:00 Oslo time (18:00 UTC). Google, Google DeepMind, the Gemini app and CEO Sundar Pichai followed up with posts on X minutes later.
A cheap model to do the heavy lifting
The Gemini 4 family now has three models. Carbon and Argon are Google's heavyweight frontier models for long, hard reasoning. Hydrogen is built for fast execution: writing code, running tests, fetching data and following instructions across many small steps.
The price is $1.50 per million input tokens and $7.50 per million output tokens. Google says the price is flat across the full one-million-token context window. The model can also produce up to one million output tokens in a single response.
Hydrogen has four thinking levels: low, medium, high (the default) and a new xhigh tier. The idea is that customers only wait and pay for as much reasoning as a task needs.
The numbers Google cites
Google says Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default level, against 50 for Argon. At the highest thinking level it reaches 54. Compared with Gemini 3.8 Flash, it is said to:
- finish tasks 2.5 times faster
- more than triple its score on the Terminal-Bench 4.1 coding test, from 18.2 to 57 percent
- make up claims more than five times less often
Early-access partners have, according to Google, seen the model run 2 to 2.8 times faster while using 12 to 23 percent fewer tokens than 3.8 Flash. These are Google's and its partners' own measurements. The coding tool Cursor says on X that Hydrogen is now available there and scores 54.3 percent on its own CursorBench test, up from 39.6 percent for 3.8 Flash.
One model directs, another does the work
Google is also launching Orchestrator in its developer tool Google Antigravity. The feature switches automatically between Carbon and Hydrogen within the same session. The heavier model plans and coordinates, while Hydrogen runs the worker and verifier agents.
Google says that on Terminal-Bench 4.0, Orchestrator came close to Carbon's success rate on its own while using 24 percent less quota on average and finishing 19 percent faster. Antigravity access applies to paying Google AI Pro and Ultra subscribers.
The biggest model switch in Google Search
Google calls this the biggest model upgrade to Search so far. According to Google, Hydrogen is available in Search for all users globally. Search is meant to understand multi-part questions, cope with typos and answer with more tables, images and links to a wider range of websites.
In the Gemini app, the model is meant to stick more closely to tone and format requirements and pick the right tool more often. It can pull context from Gmail, Google Drive and connected apps such as Airtable and Dropbox.
What it means for business leaders
- The price makes agents cheaper to run. Many AI agents use an expensive model for everything. Google's approach is to let a cheap model do most of the work and only call in the frontier model when needed. It is worth testing against your own bills.
- Benchmarks are not your workflow. The numbers come from Google and its partners. Run the model on your own tasks before switching.
- Search is changing for everyone. When Google Search answers with tables and summaries, visibility in AI answers matters more for companies that depend on being found.
- Not stated. The announcement does not say where data is processed or whether the model comes with separate terms for European business customers.
Sources and media
- Google: Introducing Gemini 4 Hydrogen
- Google: Gemini 4 Hydrogen for developers
- Google: Google Search gets a major upgrade with Gemini 4 Hydrogen
- Google DeepMind on X, 12 November 2026
- Cursor on X, 12 November 2026
Illustration: AI-generated with Grok Bot / hogby.ai