AI Sunday
EN

Google Search switches to Gemini 4 Hydrogen worldwide

2–3 minutes
Google Search passe à Gemini 4 Hydrogen dans le monde entier

Google announces that Gemini 4 Hydrogen is now available in Google Search for all users, worldwide. The model is meant to make answers more visual and better suited to complex questions.

What’s happening

According to Google’s press release, Search better understands multi-part questions, including those with typos or very precise requests. The model follows instructions better, calls tools more accurately and answers each part of the question.

The release also describes parallel search: Search runs simultaneous queries on the relevant sub-topics, which should surface more links than before. Google says Hydrogen’s reasoning strengthens its so-called “query fan-out” technique.

Answers now include more images, structured responses, comparison tables and custom tools. The release specifies that this applies in all the languages supported by Search’s AI mode.

Key figures

  • Model price, according to the dedicated page: $1.50 per million input tokens and $7.50 for output, on a 1-million-token window.
  • Score of 51 at the default setting, against 50 for the comparison model, and 54 at the highest setting.
  • Tasks finished 2.5 times faster than Gemini 3.8 Flash at the default setting.
  • Made-up claims divided by more than five, again compared with Gemini 3.8 Flash.
  • With Orchestrator, which switches between models within a single session: 24% less quota and 19% more speed, for a success rate close to Carbon.

In context

Gemini 4 Hydrogen was first presented for code and agent workflows, as we described in an article from November 12. The model page quotes its performance for software development, not for search. It also presents Hydrogen as a major upgrade for teams moving from Gemini 3.8 Flash.

Search’s release does not say whether the API price applies to search, or whether Search usage is billed.

What we don’t know yet

  • The release gives no speed or usage figure for Search. The numbers cited concern the model, not the search engine.
  • The exact list of languages is not specified, only “all languages” supported by AI mode.
  • It gives no detailed rollout timetable, nor any precise date beyond the announced availability.
  • The name of the benchmark and of the comparison model do not appear in the version received of the model page, where several sentences are cut off.
  • The custom tools mentioned are not described: the release does not say which ones, or how to use them.

This news brief was written automatically by an AI system, for greater responsiveness and to bring you ever fresher news.

Sources

À découvrir aussi sur AI Sunday

Portrait de Daniel, fondateur d'AI Sunday

Écrit par

Daniel

Professionnel des achats, j'utilise l'IA au quotidien dans mon travail et je raconte ici ce qui marche vraiment, et ce qui coince. En savoir plus

Commentaires

Leave a Reply