Google swaps in Gemini 4 Hydrogen for 3.8 Flash
Most people in Google's ecosystem use its Flash-class models. That is the role of Gemini 4 Hydrogen, which has taken over from Gemini 3.8 Flash across the company's products.
What Google says Hydrogen does better
Hydrogen runs at one of four thinking levels: low, medium, high and a new one called xhigh. Google says that even at high it is 2.5 times faster than 3.8 Flash.
The company also reports that its own analysis shows more than five times fewer "fabricated claims." Code generation improves as well, with Hydrogen scoring more than triple 3.8 Flash on Terminal-Bench.
Against Argon, one of Google's recent higher-end models, it sometimes holds its own. On the Artificial Analysis Intelligence Index, Hydrogen in xhigh mode scored 54, while Argon scored 50.
Faster access than Argon and Carbon
Argon and Carbon rolled out in stages, starting with cybersecurity researchers, and even their general release went only to subscribers of the priciest AI plan. Hydrogen skips that staging: it was available at once, with looser limits.
Search gets the new model
Search is the other big target: Google calls the Hydrogen integration the largest update ever to its AI-powered search. Visual elements will appear more often in AI Overviews and AI Mode answers, and the model is aimed at long questions with several parts. Google also says Hydrogen's fan-out search now cites a wider range of sources, though the company has made that claim before.
Everyone is reportedly getting Hydrogen in search. In the Gemini app and AI Studio, access is tighter. The model is live for AI Pro and Ultra subscribers, while AI Plus and free users remain on the older 3.5 Flash-Lite.
Developer pricing and coding use
Through the API, Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens. That is unchanged from 3.8 Flash. Carbon and Argon cost $2 and $10 for the same amounts today, and they rise to $4 and $20 starting next month.
For agentic coding, Google sees Hydrogen as a central tool. Argon and Carbon remain its pick for long-horizon tasks and security work, while Hydrogen's speed is aimed at simpler code and test runs that would otherwise burn through token budgets.
Source: Ars Technica