Google releases Gemini 4 Hydrogen at $1.50 in and $7.50 out per million tokens

Google released Gemini 4 Hydrogen today. Google calls it the company's "daily driver for agentic workflows". The model is generally available now. It is in AI Studio, the Gemini API, Vertex AI and the Gemini app.

Google's list price is $1.50 per million input tokens. The output price is $7.50 per million tokens. The price is flat across the whole context window. The window holds one million input tokens. It can return up to one million output tokens.

What Google claims about scores

Google places Hydrogen at 51 on the Artificial Analysis Intelligence Index. The score uses the model's default thinking level. Google places Argon, its larger sibling, at 50 on the same index. Both figures are Google's own in the announcement.

Google says Hydrogen outperforms Argon on Terminal-Bench 4.0. Google says it ranks first on Vals Tax Agent Bench and first on Vals Legal Research Bench. On Agent Arena, Google says Hydrogen ranks fifth. These are the vendor's own scores at launch.

The four thinking levels

Hydrogen is offered at four thinking levels: low, medium, high and xhigh. The default is high. Google says the levels give "scalable quality, latency and cost tradeoffs". A lower level returns an answer faster, at a lower cost. A higher level costs more time and more tokens.

Google says the model handles iterative coding and debugging with "surgical edits". It handles multi-step agentic execution. It also reads long PDFs, spreadsheets, audio and video. Google says it can draft web pages and UI mockups from a prompt.

Where it sits in the Gemini line

Argon still exists. Google's Argon landing page lists Argon at $2 per million input tokens. Argon's output is $10 per million tokens, at its standard tier. Hydrogen is 25 percent cheaper per million input tokens than Argon. It is 25 percent cheaper per million output tokens than Argon. All prices are Google's own list prices.

Flash remains the lower tier. Google places Hydrogen between Flash and Argon. Hydrogen is the model to reach for when Flash is too limited and Argon is too expensive. Google says Hydrogen is "faster than models in its class".

Consumer access rolls out at the same time. The Gemini app is switching to Hydrogen for all users globally, per Google. Google Search is also moving to Hydrogen in a separate announcement.

A team paying Argon prices on most calls has a reason to test the same task on Hydrogen. Read the eval score on your own task. Google's figures say the quality drop is one point on the intelligence index. The saving is 50 cents per million input tokens and $2.50 per million output tokens. On a six-figure monthly model bill, the arithmetic is the whole decision. Measure before you move the default.

Source

Introducing Gemini 4 Hydrogen, Google.