Google releases Gemini 4 Hydrogen, a faster, cheaper AI model now powering Google Search and the Gemini app for everyone
Google says the new model matches its bigger Gemini 4 models on one widely watched test while finishing tasks faster, and makes up facts far less often than the model it replaces.
Google on Nov. 12 released Gemini 4 Hydrogen, a faster and cheaper version of its newest artificial intelligence model, and began putting it into Google Search for all users worldwide and into the Gemini chatbot app.
For most people, that means the AI answers at the top of some Google searches and many Gemini app replies will now come from the new model. Free Gemini app users get Hydrogen automatically when the app judges it right for a question, and paying subscribers can pick it from a menu, Google said.
What changes for users
- Search should better understand long, multi-part questions and typos, give more visual answers and link to a wider range of websites, Google said.
- In the Gemini app, Google says the model follows instructions about tone and format more closely and more often uses the right tools, such as pulling information from Gmail or Google Drive when a user allows it.
- Google says Hydrogen makes up false claims more than five times less often than Gemini 3.8 Flash, the fast model it replaces, and finishes tasks 2.5 times faster.
What the model is
An AI model is the underlying program that writes the answers in a chatbot like Gemini, ChatGPT or Claude. Companies usually offer large, slower models for hard problems and smaller, faster ones for everyday work. Google’s two larger new models, called Gemini 4 Carbon and Gemini 4 Argon, are meant for long, difficult reasoning. Hydrogen is the quick one, built for routine tasks and for “agents,” AI programs that carry out multistep jobs such as writing and testing computer code.
Google says Hydrogen slightly outscored Argon on the Artificial Analysis Intelligence Index, a combined score from an independent firm that tests AI models, 51 to 50 at its default setting. Those figures come from Google and were not independently checked for this story.
For developers and businesses
Developers pay $1.50 per million “tokens” of text sent to the model and $7.50 per million tokens it writes back. A token is a small chunk of text, often part of a word. The model can take in about 1 million tokens at once, enough for long documents or recordings. Google also added a feature called Orchestrator to Antigravity, its AI coding tool, which switches between Carbon and Hydrogen in the same session to save time and cost.
Source: Google, “Introducing Gemini 4 Hydrogen,” Koray Kavukcuoglu, Nov. 12, 2026.