Gemini 4 Hydrogen: pricing, speed and how to use it

Gemini 4 Hydrogen is Google's new fast, lower-cost model for coding and agent work, released November 12. It costs $1.50 per million input tokens and $7.50 per million output tokens, and you can try it today in the Gemini app, Google AI Studio and Google Search.

@GoogleDeepMind introduced it as "built for rapid coding, multimodal synthesis, and everyday workflows." The same post makes two claims: Hydrogen finishes tasks 2.5 times faster than Gemini 3.8 Flash, and it scores above Gemini 4 Argon on the Artificial Analysis Intelligence Index.

Argon was Google's flagship six weeks ago. We covered it in our Gemini 4 Argon launch post. Now the budget model in the same family edges past it.

What Gemini 4 Hydrogen is

In Google's announcement, Koray Kavukcuoglu of Google DeepMind calls Hydrogen "our daily driver for agentic workflows." The top models, Argon and Gemini 4 Carbon, stay in the lineup for long, hard reasoning jobs. Hydrogen is the one Google expects you to use for most everyday work.

The numbers below come from Google's post, so read them as the company's own claims:

Google also says partner tests ran 2 to 2.8 times faster than 3.8 Flash and used 12% to 23% fewer tokens.

The model has four thinking levels: low, medium, high and a new xhigh tier. High is the default. Google says it takes PDFs, spreadsheets, charts, audio and recorded video as input, with a 1 million-token context window.

Gemini 4 Hydrogen pricing

Google's post gives one flat rate across the whole context window:

For comparison, Argon and Carbon launched at $2 input and $10 output as an introductory rate, with a standard rate of $4 and $20. Hydrogen is a quarter cheaper than that promotion and well under half the standard price.

It looks less cheap next to the model it replaces. Gemini 3.8 Flash launched at 75 cents and $3.75 as an introductory price, and Google listed its standard rate at double that, as we noted in our 3.8 Flash post. Double is $1.50 and $7.50. So if you built on the Flash promotion, moving to Hydrogen doubles your per-token price. Google's claim of lower token use offsets some of that, but even the best case in its partner tests, 23% fewer tokens, leaves you paying about half again as much per job.

Google's post doesn't mention cached-input discounts, batch pricing, rate limits or a free API allowance. It also doesn't give a shutdown date for 3.8 Flash.

How to use Gemini 4 Hydrogen

Google says the rollout is global and starts today. Where to find it:

• In the Gemini app, pick Hydrogen in the model drop-down. Free users get it through automatic model selection when the prompt suits it.
• In Google AI Studio and the Gemini API, it's available on all AI Studio tiers. The model name is gemini-4-hydrogen.
• Google Search already runs it for all users globally, so you don't have to switch anything on.
• Google Antigravity has it for paid Google AI Pro and Ultra subscribers.
• Gemini in Google Sheets gets it too, starting with Google AI Pro and Ultra subscribers.

If you want a quick test, open AI Studio, choose Hydrogen and give it a job you already know the answer to. Upload last month's supplier invoices as PDFs and ask for a table of vendor, amount and due date. Then run the same prompt at low and at xhigh, and compare the accuracy and the wait. That tells you which thinking level your work needs, and the level drives most of the cost.

Developers have one more option. Google describes an Orchestrator setup in Antigravity that pairs Carbon with Hydrogen. On Terminal-Bench 4.0, Google says it came close to Carbon's success rate while using 24% less quota and finishing 19% faster.

My read

I care more about the speed claim than the index score. One point over Argon is inside the noise, and Google chose which benchmarks to show. A model that finishes in under half the time is something you notice, though, especially when an agent chains 30 steps and you're waiting on each one.

I'd hold off on the "beats Argon" framing until independent cost-per-task numbers come out. With 3.8 Flash, the price per token stayed flat while Artificial Analysis measured the cost per task about 40% higher. With Hydrogen the per-token price went up for anyone on the Flash discount, and how far your bill follows depends on how many tokens it spends on your tasks.

What this means for businesses adopting AI

The cheap tier is now good enough for most office work: reading documents, filling in spreadsheets, drafting replies. The expensive models are for the rare hard problem. If you'd like help sorting your own processes into those two piles, New Face Design offers a free process audit.