Google began rolling out Gemini 4 Hydrogen to Google Search worldwide on November 12, the company’s newest AI model built to answer complex, multi part questions. Search now splits a single question into several simultaneous queries and can show more images, tables and relevant links in its responses.
Google confirmed the rollout in two posts on its company blog on November 12. One came from Robby Stein, Vice President of Product for Google Search, and described it as the company’s biggest model upgrade to Search to date. The other, from Koray Kavukcuoglu, Senior Vice President of Google DeepMind and Google’s Chief AI Architect, introduced Hydrogen itself as a standalone model now available across several Google products, not just Search.
Hydrogen sits below Google’s two “frontier” models, Gemini 4 Carbon and Gemini 4 Argon, which the company reserves for deep, long horizon reasoning on harder problems. Google built Hydrogen instead for fast execution across agentic workflows, tasks where an AI system carries out several steps on its own, and says it finishes tasks 2.5 times faster than Gemini 3.8 Flash, the model it now replaces as Search’s engine.
What actually changes when you search
Search can now follow longer, messier questions without losing the thread. Google’s post says Hydrogen understands multi part questions, highly specific requests and hard to read typos more precisely than before, so a question typed the way a person actually thinks it, rather than as a tidy keyword string, should return a more complete answer.
Behind that is a technique Google calls query fan-out: instead of running one search for a question, Hydrogen breaks it into several related queries at once and pulls in results from all of them. Google says this surfaces websites a user might not have found with a single search. Results also lean more visual now, with more images, structured comparison tables and interactive tools appearing directly in the answer, and the improvements apply across every language supported by Google’s AI Mode, not only English.
How Google says Hydrogen measures up
Google’s own numbers put Hydrogen ahead of its larger sibling model on at least one general intelligence test. On the Artificial Analysis Intelligence Index, Hydrogen scores 51 at its default “high” thinking setting against Argon’s 50, and reaches 54 at a new, higher “xhigh” setting. On Agent Arena, an independent live benchmark for multi step tasks such as deep research and data analysis, Google says Hydrogen ranks fifth.
Hydrogen is priced flat across a full 1 million token context window, at $1.50 per million input tokens and $7.50 per million output tokens, according to Google’s post for developers. That pricing applies to the Gemini API and Google AI Studio, not to ordinary Search use, which stays free for users as before.
Why Search runs on the smaller model, not the biggest one
Google’s argument is that Search does not need its most powerful model, it needs its fastest reliable one. Carbon and Argon are built for the hardest reasoning problems, the kind that can take longer to work through. Hydrogen trades some of that depth for speed, which matters for a product like Search where people expect an answer in a second or two, not after a long reasoning pause.
Kavukcuoglu, introducing the model, framed it as a deliberate trade off rather than a lesser release: “Hydrogen brings cutting-edge intelligence to everyone,” he wrote in Google’s announcement. The company is positioning Hydrogen as the model most people will actually interact with day to day, through Search, the Gemini app and everyday coding tools, while Carbon and Argon stay in the background for harder, slower jobs.
Where else Hydrogen shows up
Search is not the only place getting the new model. Google says Hydrogen is live today across the Gemini app, every tier of Google AI Studio, the Gemini API, Gemini for business, and for paid Google AI Pro and Ultra subscribers inside Google Antigravity, the company’s coding-focused workspace.
| Model | Google’s role for it | Where it runs today | What Google says about it |
|---|---|---|---|
| Gemini 4 Hydrogen | Fast execution for everyday and agentic tasks | Google Search, Gemini app, AI Studio, Gemini API, Antigravity | Scores 51 on the Artificial Analysis Intelligence Index at its default setting, 54 at “xhigh”; ranks 5th on Agent Arena |
| Gemini 4 Argon | Frontier model for deep reasoning | Antigravity, developer and enterprise workflows | Scores 50 on the Artificial Analysis Intelligence Index at its default setting |
| Gemini 4 Carbon | Frontier model, long horizon reasoning | Antigravity’s Orchestrator mode, enterprise workflows | Positioned by Google as the model for the hardest, slowest problems |
| Gemini 3.8 Flash | Previous fast model | Superseded in Search by Hydrogen; still available via the Gemini API | Hydrogen completes tasks 2.5 times faster than this model, per Google |
Common questions
Is this live in India? Yes. Google’s post says the rollout is global and covers every Search user, with the richer answers available in all languages supported by AI Mode.
Do I need to do anything to use it? No. Google says the update applies automatically to Google Search; there is no separate app or setting to turn on.
Google has not said whether Hydrogen will eventually power other products beyond the ones named in its announcement, or how it plans to measure the update’s effect on everyday searches going forward.
Leave a Reply