Google has rolled out Gemini 4 Hydrogen, immediately replacing Gemini 3.8 Flash across its product ecosystem with substantial speed and accuracy improvements.
Unlike the restricted, phased rollouts of Google’s previous Argon and Carbon models, Hydrogen is available right away. The new architecture operates across four distinct thinking levels—low, medium, high, and xhigh. Google reports that even when running in high mode, Hydrogen functions 2.5 times faster than Flash 3.8 while cutting fabricated responses by more than five times. In coding evaluations, the model more than tripled the Terminal-Bench score of 3.8 Flash. It also posted an Artificial Analysis Intelligence Index score of 54 in xhigh mode, edging out Argon’s score of 50.
Google is targeting Hydrogen toward agentic workflows, positioning its quick turnaround as an efficient option for generating routine code and running test suites without exhausting token limits. Argon and Carbon remain designated for long-horizon planning and security tasks. Hydrogen is also being integrated directly into Google Search to power AI Overviews and AI Mode, introducing richer visual components and improved handling for complex, multi-part inquiries.
While Hydrogen is rolling out broadly to Search users, access within the Gemini app and AI Studio is currently restricted to AI Pro and Ultra subscribers, with free and AI Plus tiers remaining on Gemini 3.5 Flash-Lite. Developer access via API matches previous Flash pricing at $1.50 per million input tokens and $7.50 per million output tokens, contrasting with Argon and Carbon rates that are slated to rise to $4 and $20 next month.
Source: Ars Technica



