Google is bringing Gemini 4 Hydrogen to Search and other products with a focus on questions that are difficult to phrase or require several answers at once. The company says the model can break a prompt into parts, address each one, and return responses with more links and visual material.
Built for questions with several parts
Google describes Gemini 4 Hydrogen as better at handling prompts that combine multiple questions, contain hard-to-understand typos, or ask for highly specific information. Instead of treating such a prompt as one undifferentiated request, the model can separate its components and respond to each part.
That approach could help when a user has a detailed question but does not know how to divide it into simpler searches. Google’s description focuses on the model’s ability to interpret the request and organize an answer around its subtopics.
Robby Stein, VP of product at Google Search, says the model is “now the new default model in AI Mode & being used for your hardest questions in AI Overviews.” That places Gemini 4 Hydrogen in two Search experiences, with AI Overviews using it for the questions Google considers most demanding.
More links and visual answers
Google says Gemini 4 Hydrogen can conduct simultaneous queries across relevant subtopics, giving users a wider range of links in its responses. The company also says answers can include more visuals, such as images, comparison tables, graphs, and maps.
One example from Stein shows the model creating an interactive animated driving simulation. The user asked how phones determine which lane of a highway a car is on in places without cell service. The example illustrates how a response might use an interactive visual to explain a technical process, rather than relying only on text.
Google also says language understanding has improved across all supported AI Mode languages. The source does not list those languages or provide measurements for the improvement, but presents broader language understanding as part of the model’s Search update.
Where Gemini 4 Hydrogen is available
Google says Gemini 4 Hydrogen is available starting today in Search, the Gemini app, all tiers of Google AI Studio, the Gemini API, Gemini for business, and Google Antigravity for Google AI Pro and Ultra subscribers.
Availability across these products means the model is not limited to a single Search feature. Google’s announcement covers consumer-facing experiences, developer access, and business offerings, though the source does not detail how the model’s behavior may differ across them.
Thinking settings and model comparisons
Users have four thinking level options: low, medium, high, and a new “xhigh” mode. These options give users a way to select how much reasoning effort the model applies, according to the available settings described in the announcement.
Google says Gemini 4 Hydrogen delivers significant leaps in both speed and intelligence compared to Gemini Flash 3.8. The company also says it “matches or outperforms some of the competing models in its class,” naming Claude Sonnet 5.5 and GPT-6.1 Terra.
Those comparisons are Google’s claims; the source provides no benchmark results or testing details. The practical measure for users will be whether the model can follow complicated prompts, surface useful references, and choose an appropriate format for its answers.