Gemini 4 Hydrogen: Google's Fast Gemini 4 Model for Everyone
Gemini 4 Hydrogen is Google's fast Gemini 4 model: free in Search and the API, $1.50/$7.50 per 1M tokens paid. Who gets it, prices and how it compares to Argon.
PromptWises is reader-supported. Some links may earn us a commission at no extra cost to you. This never affects our ratings or recommendations.
Gemini 4 Hydrogen is the fast, cheaper model in Google’s Gemini 4 family and the first Gemini 4 model everyone can reach. Google released it on November 12, 2026, globally in Google Search, in the Gemini app, and on the free and paid tiers of the Gemini API at $1.50 per million input tokens and $7.50 per million output tokens. Since November 13 it is also in GitHub Copilot. Sources: Google’s announcement by Koray Kavukcuoglu, the Gemini API guide to Hydrogen and the Search post.
Hydrogen vs Carbon and Argon
Google now has three Gemini 4 models. Gemini 4 Carbon is the strongest and spends the most thinking tokens. Gemini 4 Argon handles most advanced long-horizon work with fewer tokens. Hydrogen, “our daily driver for agentic workflows” in Google’s words, is built for speed: everyday coding, web apps, tool-using agents and document work. It is also Google’s recommended upgrade from Gemini 3.8 Flash and older models.
All three share a 1,048,576-token window, and long decode lets a response run past the default 64K output tokens up to 1M. Hydrogen adds a fourth thinking level to the API: low, medium, high (the default) and a new xhigh for hard, multi-step problems. Google advises starting at high and keeping xhigh for tasks where high falls short.
Where you can use it
| Who | Where | Notes |
|---|---|---|
| Free Gemini app users | Auto model selection | Only when Gemini picks it for a prompt; most prompts use Flash-Lite |
| Google AI Plus, Pro, Ultra | Gemini app model menu | Pro adds Argon; Ultra adds Argon and Carbon |
| Everyone | Google Search | Live globally |
| Developers | Gemini API (free and paid tiers), AI Studio, Android Studio | Model ID gemini-4-hydrogen |
| AI Pro and Ultra subscribers | Google Antigravity | Also the default model for the Antigravity agent in Gemini Managed Agents |
| Businesses | Gemini for business, Gemini in Sheets (eligible plans) | Consumer Sheets access starts with AI Pro and Ultra |
| Copilot Pro, Pro+, Max, Business, Enterprise | GitHub Copilot model picker (since Nov 13) | Usage-based billing at provider list pricing; gradual rollout (GitHub) |
One point is easy to misread: Google’s footnote says free users get Hydrogen only “via auto model selection when it’s the appropriate model for the prompt”. Google’s model help page adds that most free prompts still use Flash-Lite. Plus subscribers now get Hydrogen in place of the older Flash model, which makes the $4.99 plan better value; our Gemini review has the updated plan table.
API pricing compared
Paid-tier prices per million tokens, at the time of writing (Gemini API pricing):
| Model | Input | Output | Free tier |
|---|---|---|---|
| Gemini 4 Hydrogen | $1.50 | $7.50 | Yes |
| Gemini 3.8 Flash | $0.75 until Dec 31, then $1.50 | $3.75 until Dec 31, then $7.50 | Yes |
| Gemini 4 Argon | $2 until Nov 30, then $4 | $10 until Nov 30, then $20 | No |
| Gemini 4 Carbon | $2 until Nov 30, then $4 | $10 until Nov 30, then $20 | No |
Hydrogen’s price is flat across the full context, cached input is $0.075 per million, and Batch and Flex are half price. 3.8 Flash’s introductory price ends on December 31, after which it costs the same as Hydrogen, so test the switch in December. GitHub Copilot removes 3.8 Flash on December 14, suggesting Hydrogen.
What changes in Google Search
Google calls Hydrogen “the biggest model upgrade to Search yet”. It lists better handling of multi-part questions, typos and very specific requests; an upgraded query fan-out that links to “a broader range of helpful links than before”; more visual answers with images, tables and custom tools; and better responses in all AI Mode languages.
The benchmark claims
Google says Hydrogen beats Argon on the Artificial Analysis Intelligence Index at its default level (51 vs 50, and 54 at xhigh), scores 57% on Terminal-Bench 4.1 against 18.2% for 3.8 Flash, and makes more than five times fewer fabricated claims than 3.8 Flash. All of these are vendor-reported. Our guide to reading AI model announcements explains how to test them.
What to do now
- Free users: keep auto model selection on (Settings > Usage limits) if you want Gemini to route harder prompts to Hydrogen.
- Developers on 3.8 Flash: test Hydrogen at
lowtohighbefore January 1 and pick the lowest level that passes. - Choosing an assistant: see our Claude vs Gemini comparison and the news section.
Frequently asked questions
Can free Gemini users use Gemini 4 Hydrogen?
Partly. Google says free users get Hydrogen through auto model selection when Gemini decides a prompt needs it; most prompts still go to Flash-Lite, and with auto selection off every answer uses Flash-Lite. Google AI Plus, Pro and Ultra subscribers can pick Hydrogen in the model menu.
How much does the Gemini 4 Hydrogen API cost?
At the time of writing, $1.50 per 1M input tokens and $7.50 per 1M output tokens (thinking tokens count as output) on the paid tier, flat across the whole 1M context. Cached input is $0.075 per 1M with implicit caching only, and Batch is half price. There is also a free tier, whose data Google may use to improve its products.
Is Gemini 4 Hydrogen better than Gemini 4 Argon?
Not for everything. Google reports that Hydrogen edges Argon on the Artificial Analysis Intelligence Index (51 vs 50), but that is a vendor-reported figure, and Google still positions Argon and Carbon for deep, long-horizon reasoning. Hydrogen is the cheaper, faster choice for everyday coding, agents and document work.
Should I move from Gemini 3.8 Flash to Hydrogen?
Google recommends it. Note the price: 3.8 Flash costs $0.75 / $3.75 per 1M tokens until December 31, 2026, then rises to $1.50 / $7.50 from January 1, 2027, the same as Hydrogen. Test Hydrogen on your own prompts before switching.
What are Hydrogen's thinking levels?
In the API there are four: low, medium, high (the default) and a new xhigh for hard, multi-step problems. In the Gemini app, Google's help page lists Low, Medium and High.