Google brings Gemini 4 Hydrogen to Search, promising a wider mix of links in its answers

Google on Thursday, November 12, announced Gemini 4 Hydrogen, an AI model aimed at coding, multimodal tasks and agentic workflows. The company also rolled the model into Google Search, changing how queries are understood and how results are displayed.

According to a report by Olhar Digital published on November 12, Hydrogen can interpret natural language searches even when a user packs several questions into a single query or makes typos. Answers will now include more visual elements and a wider variety of relevant links and websites.

In the Gemini app, Google says the model will improve everyday tasks, writing, research and integration with services such as Gmail and Google Drive. Free users may get access to the model through automatic selection, when the system decides it is the right fit for the request.

Faster, and less likely to make things up

Google says the model completes tasks up to 2.5 times faster than its predecessor, Gemini 3.8 Flash. In internal testing, the company claims it cut the number of fabricated claims by more than five times compared with the previous version.

On coding, its score on the Terminal-Bench 4.1 benchmark rose from 18.2% for Gemini 3.8 Flash to 57%. On the Artificial Analysis Intelligence Index, Hydrogen scored 51 points in its default advanced reasoning setting, against 50 for Gemini 4 Argon, and reached 54 in its maximum setting.

The model also handles PDFs, spreadsheets, charts, audio and video, and can pull information from different formats into structured reports, with references to the source materials. It offers four reasoning levels: low, medium, high and xhigh.

Working as a team, and pricing

Hydrogen can run alongside other models in the Gemini 4 family, such as Carbon and Argon. The more advanced models take on complex decisions and deep analysis, while Hydrogen handles fast execution, such as writing code, running tests and checking changes. To make this work, Google introduced Orchestrator, a feature of its Antigravity platform that switches between models automatically within the same session. On Terminal-Bench 4.0, the tool came close to the results of Gemini 4 Carbon while using 24% less quota on average and finishing tasks 19% faster.

In the Gemini API, the announced price is $1.50 per million input tokens and $7.50 per million output tokens, with support for up to one million input tokens and one million output tokens. The model is also available in Google AI Studio, Android Studio, Gemini for business and, in Antigravity, for Google AI Pro and Ultra subscribers.

What it means for brands

For brands, Search is the part that matters most. With Hydrogen, Google gets better at handling long, conversational queries that bundle several questions together, and at showing a wider variety of links and websites in its answers. That is worth factoring into your AEO and content planning: build content that answers questions asked this way.