KEY POINTS
- Google released Gemini 4 Hydrogen, which it says answers complex, multi-part questions more helpfully.
- The model is now the default in AI Mode and is used for the hardest questions in AI Overviews.
- Google says it is faster and smarter than Gemini Flash 3.8 and matches or beats some rival models in its class.
Google has introduced a new artificial intelligence model, Gemini 4 Hydrogen, and says it can give more useful answers to complicated questions. The Verge reported on the launch on November 12, 2026, and the model is available starting that day.
According to Google, the model is designed for questions with several parts, hard-to-understand typos and highly specific requests. It can split a question into its components and address each of them in its reply.
Robby Stein, Google's VP of product for Search, described the release as the biggest model upgrade to Search so far. He said it is now the default model in AI Mode and is being used for the hardest questions in AI Overviews.
Availability is broad. Gemini 4 Hydrogen is offered in Search, the Gemini app, all tiers of Google AI Studio, the Gemini API and Gemini for business. In Google Antigravity, it is available to subscribers of Google AI Pro and Ultra.
Google also says the model returns a wider range of links in its answers. It does this by running simultaneous queries across all the relevant subtopics of a question. Responses now include more visual elements such as images, comparison tables, graphs and maps.
One example shared by Stein showed the system generating an interactive animated driving simulation. The prompt was a user's question about how phones work out which lane of a highway a car is in when there is no cell service. Google also says language understanding has improved across all languages supported in AI Mode.
Users get more control over how much effort the model puts into a reply. The thinking level options are low, medium and high, plus a new mode called xhigh.
On performance, Google says Gemini 4 Hydrogen delivers significant leaps in speed and intelligence compared with Gemini Flash 3.8. The company also says it matches or outperforms some competing models in its class, naming Claude Sonnet 5.5 and GPT-6.1 Terra. Those comparisons come from Google, and the report does not include independent test results.
