Google Launches Gemini 4 Hydrogen, Default AI Model

Google launches Gemini 4 Hydrogen and activates it as the default model in AI mode. The company claims improvements in speed and intelligence compared to Gemini Flash 3.8, expanded reasoning options including an "xhigh" mode, and richer responses with links and visuals. The model is being rolled out today across the main access points of the platform.

Google announces gains over Flash 3.8 and an "xhigh" mode

According to Google, Gemini 4 Hydrogen brings significant advancements in speed and intelligence compared to Gemini Flash 3.8, and it would match or exceed some competitors like Claude Sonnet 5.5 and GPT-6.1 Terra. The model offers several levels of reasoning (low, medium, high) and adds a new "xhigh" mode. Google also indicates improved linguistic understanding across all languages that the AI mode already supports.

Immediate deployment in Search, the Gemini app, AI Studio, and the API

Gemini 4 Hydrogen is now accessible in Google Search, the Gemini app, across all offerings of Google AI Studio, via the Gemini API, as well as in Gemini for enterprises. This version is also available in Google Antigravity for users with a Google AI Pro or Ultra subscription. Robby Stein, Vice President of Product at Google Search, states that this model becomes the default in AI mode and is used for the most challenging questions in AI previews.

More structured responses, more links, and varied visuals

Google claims that Gemini 4 Hydrogen handles complex queries better, including multi-faceted ones, those with typos, or very specific requests, by breaking down each part of the question. The company specifies that it launches simultaneous queries on relevant sub-topics to provide a greater variety of links. Responses incorporate more images, comparison tables, graphs, and maps. Robby Stein notably cites an interactive animated driving simulation produced in response to a question about locating a highway lane without cellular service.