The short version: Gemini 4 Hydrogen, Google’s new everyday model, costs $1.50 per million input tokens and $7.50 per million output tokens on the Gemini API. That looks like double the $0.75 and $3.75 Google charges today for Gemini 3.8 Flash, the model it replaces. It is not, for long: Google’s own pricing page, read November 14, 2026, says Flash’s rate ends December 31 and rises on January 1 to exactly $1.50 and $7.50. From January, the newer model and the older one cost the same per token. In the free Gemini app, though, “available to everyone” has a footnote: free users get Hydrogen only when Google’s automatic model picker decides a question needs it.
That is mostly good news, and worth saying plainly. A model Google rates above its own Gemini 4 Argon on one independent index is now on the free tier of the API, in Google Search for everyone, and in the cheapest paid Gemini plan. What follows is who actually gets Gemini 4 Hydrogen, what it costs a small business that runs software on it, and two dates on Google’s pricing page that matter more than the launch.
What is Gemini 4 Hydrogen?
Google announced Hydrogen on November 12 in a post by Koray Kavukcuoglu, Google DeepMind’s senior vice president. It is the fast, everyday member of the Gemini 4 family, sitting below the heavyweight reasoning models Gemini 4 Carbon and Gemini 4 Argon. Google built it for “rapid execution” of routine work: drafting, coding, reading files, calling tools, and following instructions without wandering off.
Google’s headline claims, all its own measurements: Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default thinking level against Argon’s 50, finishes tasks 2.5 times faster than Gemini 3.8 Flash, and cuts fabricated claims by more than five times compared with Flash. Early partners, Google says, saw it use 12 to 23 percent fewer tokens than Flash on data-heavy jobs. It reads up to 1 million tokens at once, well over a thousand pages of text, and Google prices that whole window flat, with no surcharge for long documents.
Benchmarks from a launch post are a starting point, not a verdict. The useful questions for an owner are narrower: can I use it, and what will it cost the tools I already pay for?
Can free Gemini users get Gemini 4 Hydrogen?
Sometimes, and not by choice. The launch post says everyone can “start using it in the Gemini app model drop down,” but a footnote on the same page adds that free users will get Hydrogen “via auto model selection when it’s the appropriate model for the prompt.”
Google’s About Gemini models help page, read November 14, spells out what that means. With no Google AI plan, the app offers “Auto or Flash-Lite.” If auto selection is on, “most prompts will use Flash-Lite for quick answers,” and prompts that need more reasoning “may use Hydrogen.” If auto selection is off, every answer comes from Flash-Lite, which Google describes as its model for quick answers to everyday questions. A free user cannot pick Hydrogen from the menu.
Here is the access ladder from that page, with US prices from the Google One plans page, also read November 14:
| Plan | Monthly price | Hydrogen in chat |
|---|---|---|
| No plan (free) | $0 | Only when auto selection picks it |
| Google AI Plus (2 TB) | $9.99 | Yes, selectable |
| Google AI Pro (5 TB) | $19.99 | Yes, plus Argon and task mode |
One check on a number in circulation: several outlets reported in June that Google AI Plus had dropped to $4.99 a month in the US. On November 14 the Google One plans page listed Google AI Plus only as a 2 TB plan at $9.99 a month, and its description is the one that names “access to our Hydrogen model.” If you are budgeting for it, use the price on Google’s page when you sign up, not the headline from June.
Two more details from the help page. The app offers low, medium and high thinking levels, not the extra-high setting developers get, and if you do not choose one, it “will default to Low.” The help page also notes that it describes personal Google accounts; on a work Google Workspace account, which models you see “will depend on your Workspace license.” We covered why the login you use for Gemini decides whether your chats train Google’s models, and that still applies here.
How to see which model answered you
Google gives free users a way to check. On a computer at gemini.google.com, open the chat, click More under a response, choose Show thinking steps, and scroll to the bottom; the model is named there. If you rely on the free app for something that matters, such as a customer email or a summary of a contract, this is worth ten seconds. The difference between Flash-Lite and Hydrogen on a careful job is exactly the kind of difference you would want to know about.
Gemini 4 Hydrogen pricing on the API
Most small businesses never touch an API. But plenty of the tools they rent do: the reply drafter in a help desk, the call summarizer in a phone system, the quote assistant a developer wired into a website. If any of those runs on Gemini, the API price shows up eventually in what the vendor or developer charges you.
Here is what Google’s Gemini API pricing page showed on November 14, per million tokens on the paid tier:
| Model | Input now | Output now | Scheduled change |
|---|---|---|---|
| Gemini 4 Hydrogen | $1.50 | $7.50 | None listed |
| Gemini 3.8 Flash | $0.75 | $3.75 | $1.50 and $7.50 from January 1, 2027 |
| Gemini 4 Argon | $2.00 | $10.00 | $4.00 and $20.00 from December 1, 2026 |
Read across the rows and the launch price looks different. For the next seven weeks Hydrogen costs twice what Flash does. On January 1 the gap closes to zero. Argon, the step up, doubles a month sooner; we flagged that date when Gemini 4 Carbon launched, and it has not moved. Output prices include the model’s thinking tokens, so a higher thinking level costs more even when the visible answer is the same length.
Hydrogen is also on the API’s free tier, which Argon and Carbon are not. The free tier carries two conditions on the same page: content is “used to improve our products,” and grounding with Google Search is not available. Free is fine for testing with made-up data. It is the wrong place for real customer emails.
What does Hydrogen cost for a real small-business job?
Take a six-person heating and cooling company whose help desk tool drafts replies to customer emails with a Gemini model. Say 40 emails a day, every day, so 1,200 a month. Each call sends 2,000 tokens (the email, the thread and the instructions) and gets back 1,500 tokens, counting the model’s thinking. That is 2.4 million input tokens and 1.8 million output tokens a month.
- On Gemini 3.8 Flash today: 2.4 times $0.75 is $1.80, plus 1.8 times $3.75 is $6.75. Total: $8.55 a month.
- On Hydrogen today: 2.4 times $1.50 is $3.60, plus 1.8 times $7.50 is $13.50. Total: $17.10 a month.
- On Flash from January 1: the same $17.10.
- On Hydrogen, if Google’s partner results hold: 12 to 23 percent fewer tokens for the same work puts it at roughly $13.17 to $15.05 a month, below what Flash will cost for the same emails.
The dollar amounts are small either way, which is itself the point. For a job this size the model price is a rounding error next to one missed service call. The question worth asking the vendor is not “which model is cheapest” but “which model gets the reply right more often,” and Google’s claim of five times fewer fabricated claims than Flash is the number to test, with your own emails, before believing it.
The December 1 switch your old tools may already be on
The pricing page carries a second date that Google’s launch post does not mention. Effective December 1, 2026, every request to the retired Gemini 2.5 Pro model will be sent automatically to Gemini 3.8 Flash and billed at Flash rates, and every request to Gemini 2.5 Flash will be sent to Gemini 3.5 Flash-Lite and billed at Flash-Lite rates.
For a tool built a year or two ago and never touched since, that means the model underneath changes on December 1 whether anyone at the vendor does anything or not. On price alone it is good news: 2.5 Pro cost $1.25 and $10.00 per million tokens for ordinary prompts, so moving to Flash is cheaper, at least until Flash’s own price doubles on January 1. On behavior it is unknown. A prompt tuned for one model can read differently on another, and a quote or reply tool that worked fine in November could start phrasing things oddly in December.
If a developer or vendor built something for you on Gemini, one short email covers it: which Gemini model does it call, what happens to it on December 1 and January 1, and will they test it on Hydrogen?
Why the Search change matters to your website
Hydrogen also now powers Google Search for all users. Google’s post says it upgrades the “query fan-out” technique, where Search breaks one question into many smaller searches behind the scenes, to connect people “to a broader range of helpful links than before” and to sites “you previously might not have discovered.”
That is Google’s description, not a measured result, and nobody outside Google can yet say how many more sites get shown. But the direction favors small businesses with pages that answer specific questions well: the service-area page that names the neighborhoods, the pricing page that actually lists prices, the FAQ written in customers’ own words. If Search is now asking more sub-questions per query, there are more chances for a narrow, useful page to be the answer to one of them. Our earlier look at Google’s Decisions API showed the same pattern from the other side: Google’s AI is getting better at specific, bounded questions, which is where a small business’s knowledge lives.
Frequently Asked Questions
Is Gemini 4 Hydrogen free?
Partly. It is on the Gemini API’s free tier, where Google uses your content to improve its products, and it powers Google Search for everyone. In the free Gemini app you cannot choose it; Google’s automatic model picker uses it only for prompts it judges need more reasoning, and most prompts go to Flash-Lite. Choosing Hydrogen yourself takes a Google AI Plus plan or higher.
How much does Gemini 4 Hydrogen cost on the API?
On November 14, 2026, Google’s pricing page listed $1.50 per million input tokens and $7.50 per million output tokens on the paid tier, including thinking tokens, flat across the full 1 million-token window. Gemini 3.8 Flash costs half that until December 31 and rises to the same $1.50 and $7.50 on January 1, 2027.
Should a small business switch tools to Hydrogen now?
Not in a hurry. For the next seven weeks it costs twice as much as Flash per token, and from January the two cost the same, so the sensible move is to ask your vendor or developer to test Hydrogen on your real work now and switch when it proves more accurate. If Google’s claims of fewer fabricated answers hold on your own emails or quotes, that matters far more than a few dollars a month.
What happens to tools built on Gemini 2.5 on December 1?
Google’s pricing page says requests to Gemini 2.5 Pro will be routed automatically to Gemini 3.8 Flash and billed at Flash rates, and requests to Gemini 2.5 Flash will go to Gemini 3.5 Flash-Lite at Flash-Lite rates. The tool keeps working, but the model answering changes, so ask whoever built it to check the output before that date.
Have you caught the free Gemini app switching models on you mid-conversation, and did the answer get better or worse? Tell us in the comments.
