Google has a new model, and it is going everywhere at once. Gemini 4 Hydrogen was announced on 12 November 2026 and is available from launch in Search, the Gemini app, all tiers of Google AI Studio, the Gemini API, Gemini for business, and Google Antigravity for Google AI Pro and Ultra subscribers S1.
What Google says it does
According to The Verge, Google says the model gives more helpful answers to complicated questions, naming "questions with multiple parts, hard-to-understand typos, or highly specific requests" as the target S1. The stated improvement is structural: the model breaks a question into its parts and addresses each one in its answer S1.
That is a meaningful shift for search behaviour. People rarely type one clean question. They stack a comparison, a constraint and a follow-up into a single box. A model that can separate those pieces and answer them individually is aimed squarely at that habit.
Where it lands
The rollout covers consumer and commercial surfaces at the same time S1. Robby Stein, vice president of product at Google Search, said the model is "now the new default model in AI Mode & being used for your hardest questions in AI Overviews" S1.
The money angle
The distribution list is the story. Google is putting its newest model in front of search users, the Gemini app, paying developer customers and business accounts simultaneously, rather than staging the release product by product S1. For developers, access across all tiers of AI Studio is notable: the new model is not held back to a single top plan at launch S1. For subscribers, the model reaches Google Antigravity as part of the existing AI Pro and Ultra tiers S1.
The source does not disclose what any of this costs. There is no per-token API pricing, no word of subscription price changes, and no compute or data-centre figures attached to the launch S1. Nor does it include benchmark results beyond Google's own description of the model's abilities S1. For anyone modelling AI spend, those gaps matter more than the feature list.
What to watch
First, independent testing. Google's claims about multi-part questions and typos come from Google, and third-party evaluations were not part of the material reviewed here S1. Second, pricing. Whether API rates for the new default model shift is unconfirmed S1. Third, how quickly the model becomes the default rather than an option across each of the named surfaces, since Google has confirmed default status only for AI Mode search S1.
Gemini 4 Hydrogen is not a standalone product launch. It is a replacement, quietly pushed into the layers where most people and most businesses already touch Google's AI.
- Google’s new Gemini model is better at understanding complicated questions The Verge · theverge.com