Google says its new Gemini 4 Hydrogen mannequin finishes duties 2.5 occasions sooner than the mannequin it replaces. Since Thursday it has been the primary mannequin behind AI solutions in Google Search, so most individuals who search will meet it with out selecting it.
The new model was announced in a company blog post on 12 November. Gemini 4 Hydrogen is the quick, cheaper member of the Gemini 4 household. It sits under Gemini 4 Carbon and Gemini 4 Argon, the bigger fashions Google goals on the hardest issues. It takes over from Gemini 3.8 Flash, which got here out in September.
What modifications for individuals who use Google
In Search, Gemini 4 Hydrogen is now the default mannequin for AI Mode, the chat-style search web page. Google additionally makes use of it for the toughest questions in AI Overviews, the summaries on the high of outcomes. The corporate says it copes higher with lengthy questions, typos and questions in a number of components, and that it reveals a wider vary of hyperlinks to web sites.
Within the Gemini app, paying customers can select the brand new mannequin from a listing of fashions. Free customers get the brand new mannequin solely when the app’s computerized setting picks it for a query.
The numbers Google offers
All of those figures come from Google and haven’t been checked by anybody else. Google says the mannequin scores 57% on Terminal-Bench 4.1, a take a look at of coding duties performed in a pc terminal, towards 18.2% for the older mannequin. It additionally says the brand new mannequin makes up information greater than 5 occasions much less typically.
Synthetic Evaluation, an unbiased group, runs a rating of AI fashions referred to as the Intelligence Index. Google says the brand new mannequin scores 51 on this index at its default setting. That’s one level greater than Gemini 4 Argon, the bigger mannequin. At its highest reasoning setting, the rating rises to 54.
For builders, it prices $1.50 per million tokens of enter and $7.50 per million tokens of output. A token is a small piece of textual content, typically a part of a phrase. There is also a free tier for developers, with limits on use.
What Google says about security
Google says the brand new mannequin is as much as two occasions much less prone to get round permission limits when a process is blocked. The model card lists known limits and safety test results. Just like the pace figures, these are Google’s personal take a look at outcomes.
The launch comes as AI corporations compete on worth and pace. Anthropic made its Claude Fable 5.2 mannequin 20% cheaper this week. We reported on that price cut on Tuesday. Earlier this month, OpenAI mentioned ChatGPT now runs a couple of fifth as many searches as Google. We covered those figures in a separate article.
Why it issues: Search is the place most individuals meet AI with out asking for it. A change to the mannequin behind AI Mode modifications the solutions {that a} very giant variety of folks see. Thus far, Google has not printed unbiased checks of its claims.

