Google's Gemini 4 Hydrogen tackles complex questions

Google is positioning Gemini 4 Hydrogen as a step up for handling messy, real-world queries. According to Robby Stein, VP of product at Google Search, it is now the default model in AI Mode and is being used for the hardest questions in AI Overviews. That suggests Google sees the model as core to how its search products answer complex user requests, not just a developer-only release.

The model's improvements go beyond text understanding. Google says it gives a wider range of links by conducting simultaneous queries across relevant subtopics, and its responses include more visuals like images, comparison tables, graphs, and maps. One example Stein shared shows the AI generating an interactive animated driving simulation in response to a question about how phones locate a car's lane without cell service. Such examples hint at a push toward richer, more interactive answers, though how often users encounter them in practice is not specified.

Users also get more control over thinking levels, with low, medium, high, and a new xhigh mode. Google claims significant leaps in speed and intelligence compared to Gemini Flash 3.8, and says the model matches or outperforms some competitors like Claude Sonnet 5.5 and GPT-6.1 Terra. The broad availability across search, the Gemini app, and developer tools means many users can try it immediately, and their experience will show whether these claims hold up in everyday use.