Google has put Gemini 4 Hydrogen into Google Search for all users worldwide, calling it the biggest model upgrade to Search yet. The same model also reaches the Gemini app’s model dropdown and the Gemini API on November 12, 2026.
If you type long, messy, multi-part questions into Search, this is the change you are most likely to feel, and no paid plan is needed for that part. The Gemini app is more selective, which is where the confusion will start.
TL;DR: Gemini 4 Hydrogen is a fast, cost-efficient Google model that now powers Search for everyone, with no staged schedule given. In the Gemini app, free users only get it when Google’s auto selection decides a prompt needs it, and a footnote says most prompts will still use Flash-Lite. Developers pay $1.50 per million input tokens and $7.50 per million output tokens. The benchmark numbers are Google’s own.
Gemini 4 Hydrogen in Google Search: what changes
Google’s Search post, written by Robby Stein, says Hydrogen is available globally in Search for all users, across every supported AI Mode language. Google lists four upgrades: better understanding of multi-part or typo-filled questions, wider query fan-out across subtopics, more visual answers, and more nuanced replies in each language.
The visual part is the one to watch. Google says answers can now include images, structured responses and comparison tables, so a question that weighs two phones on battery, camera and price may come back as a table instead of a wall of text.
Where else the model shows up depends on who you are, and that is the next question.
| Where | Who gets Hydrogen | Detail |
|---|---|---|
| Google Search | All users | Global, all supported AI Mode languages |
| Gemini app | Model dropdown, plus free users via auto selection | Free users get it only when judged appropriate |
| Gemini API and Google AI Studio | Developers on all tiers | Paid tier in Android Studio |
| Google Antigravity | Paid AI Pro and Ultra subscribers | Orchestrator pairs it with Gemini 4 Carbon |
What Hydrogen costs and how Google says it performs
Google’s Gemini 4 Hydrogen announcement sets pricing at $1.50 per million input tokens and $7.50 per million output tokens, flat across the full context window. That window is 1 million tokens of input, with up to 1 million tokens of output.
On speed, Google says Hydrogen finishes tasks about 2.5 times faster than Gemini 3.8 Flash at its default high thinking level. It also scores 51 on the Artificial Analysis Intelligence Index at high, against 50 for Gemini 4 Argon, and 54 at the new xhigh tier.
A fast model edging past a frontier one on a single index is a result worth treating as Google’s claim until independent testing catches up. Google also reports 57% on Terminal-Bench 4.1 against 18.2% for Gemini 3.8 Flash, and every figure here comes from Google’s own reporting.
Numbers like these matter less to most people than the question of whether their own chat window changes.
What free Gemini app users should expect
Free users in the Gemini app get Hydrogen through auto model selection, and only when Google judges it appropriate for the prompt. A footnote in the announcement says most prompts will still run on Flash-Lite.
So if you are on the free plan and nothing feels different tomorrow, that is how the system is designed. The post does not state rate limits or usage caps, which leaves the free tier’s real ceiling unknown for now.
Anyone who wants to choose the model on purpose has to use the dropdown, and that choice sits alongside two frontier siblings.
Where Hydrogen sits next to Carbon and Argon
Google describes Gemini 4 Carbon and Argon as the frontier models for deep, long-horizon reasoning, while Hydrogen is built for rapid execution. It can run on its own or act as the execution engine beside them.
In Antigravity, Google says the Orchestrator setup reaches near-Carbon task success while using about 24% less quota and finishing 19% faster on average. We covered who can use Gemini 4 Carbon and how Gemini 4 Argon reached paid Gemini app plans earlier, and Hydrogen is the faster execution layer beside both.
For most readers the practical effect is small and mostly invisible: Search answers get richer, and the Gemini app quietly picks a model for you. Our AI model tracker keeps the current list of Gemini, Claude and GPT models in one place.
