Google launches Gemini 4 Hydrogen to replace Gemini 3.8 Flash — Ars Technica

Google has begun rolling out the Gemini 4 Hydrogen AI model, which has replaced Gemini 3.8 Flash in the company’s products. The new model is designed for fast and cost-efficient tasks previously handled by models in the Flash line, Ars Technica reports.

Speed and reasoning levels

Gemini 4 Hydrogen has four reasoning levels: low, medium, high, and the new xhigh. According to Google’s assessment, in high reasoning mode, Hydrogen operates 2.5 times faster than Gemini 3.8 Flash. The company also said that the model reduces the number of fabricated claims by more than five times.

Google reports improvements in code generation: in the Terminal-Bench test, Hydrogen scored more than three times higher than Gemini 3.8 Flash. In xhigh mode, the model received 54 points in the Artificial Analysis Intelligence Index, while Argon has 50 points. At the same time, Google positions the Argon and Carbon models as better suited for long-running tasks and security.

More current news is available on the UA.News Telegram channel Telegram.

Search integration and developer access

Hydrogen will be integrated into Google Search. The company calls this the largest update to its AI-powered search tools. According to Google, the model should better process long, multi-part queries, while answers in AI Overviews and AI Mode will contain more visual elements. Google also claims that the fan-out search mechanism is configured to cite more sources.

All users are expected to receive access to Hydrogen in Search. In the Gemini app and AI Studio, the model is already available to AI Pro and Ultra subscribers, while AI Plus users and users of the free version currently use Gemini 3.5 Flash-Lite. For developers, API access costs $1.50 per million input tokens and $7.50 per million output tokens—the same price as Gemini 3.8 Flash.