Gemini 4 Hydrogen Scores 54 on Artificial Analysis, Then Uses 304 Million Tokens to Get There

Google released Gemini 4 Hydrogen on Nov. 12 and priced it at $1.50 per million input tokens and $7.50 per million output tokens. Artificial Analysis, an independent benchmarking site, lists it at 54 on its Intelligence Index at the highest effort setting, and also records that it generated 304 million tokens to finish the index.

Koray Kavukcuoglu, SVP at Google DeepMind and Chief AI Architect at Google, wrote in the launch post that Hydrogen is "our daily driver for agentic workflows". It sits beside the larger Gemini 4 Carbon and Gemini 4 Argon, which Google assigns to long-horizon reasoning.

What Google says Gemini 4 Hydrogen does

The context window is 1 million input tokens, with output of up to 1 million tokens. The price is flat across that window. Thinking levels are low, medium, high (the default) and a new xhigh.

Google did not disclose a parameter count, and the page does not state a licence. Artificial Analysis lists the model as proprietary.

Google's vendor-supplied figures include 57% on Terminal-Bench 4.1 against 18.2% for Gemini 3.8 Flash, and 51 on the Artificial Analysis index at high effort against 50 for Argon. The Google blog attributes the Orchestrator result to Terminal-Bench 4.0, while the Antigravity post says 4.1, so the version behind that figure is unclear.

What the independent page records

Measure	Hydrogen (xhigh)	Median, comparable models
Intelligence Index	54	41
Time to first answer token	41.11 s	13.45 s
Minutes per index task	13.0	12.9
Tokens generated on the index	304M	72.8M

The Artificial Analysis model page also lists an output speed of about 198 tokens per second, a blended price of $1.10 per million tokens and about $3.16 per index task. Google's headline of a model that is about 2.5 times faster than Gemini 3.8 Flash at high effort is a vendor figure. The independent page does not test that comparison.

For scale, Artificial Analysis put the larger Gemini 4 Carbon at 59 on Oct. 28, at $4 input and $20 output per million tokens. That means Hydrogen at xhigh is five points behind at well under half the list price, though the Hydrogen page does not say which index version its 54 comes from.

Where it ships

Hydrogen is rolling out globally from Nov. 12 through the Gemini API, Google AI Studio, Android Studio, the Gemini app and Google Search. In Antigravity it reaches paid AI Pro and Ultra subscribers. A new Orchestrator option, for Ultra subscribers, pairs Hydrogen with Carbon: Carbon plans read-only, Hydrogen edits and runs tests. Google reports about 24% less quota and 19% faster completion at near-Carbon success.

Agent products have carried unclear status before: see the Databricks Genie preview and the United rule banning AI agents. Hydrogen's Orchestrator is rolling out to Ultra subscribers only.

Cost per index task on the Artificial Analysis page is about $3.16, which is the figure to set against the $1.50 and $7.50 list prices. Google links a model card from the launch post and has not published Carbon's Antigravity quota figures.