Google is rolling out the Gemini 4 Hydrogen model to Search users globally to improve how the platform handles complex, multi-part questions and typos. The company described the update as its "biggest model improvement to Search yet."
Designed for fast execution across agentic workflows, Hydrogen is intended to be a faster and more efficient alternative to the company's frontier models, Gemini 4 Carbon and Gemini 4 Argon. Google reported that Hydrogen completes tasks 2.5 times faster than the previous Gemini 3.8 Flash model. Despite its speed, the model outperformed the more powerful Gemini 4 Argon on the Artificial Analysis Intelligence Index at its default high thinking level.
The model introduces four thinking levels—low, medium, high, and a new extra-high tier labeled "xhigh"—all aimed at completing tasks with fewer output tokens. On benchmarks for multi-step tasks such as data analysis and deep research, Hydrogen ranks No. 5 on Agent Arena.
In Search, the update improves user intent recognition, allowing for more nuanced queries and better navigation of typos. The model also enhances the platform's fan-out technique, which breaks single prompts into multiple subqueries to help users discover relevant websites. Users in AI Mode can expect more visual responses, including comparison tables, structured responses, and custom tools.
Beyond Search, Gemini 4 Hydrogen is being released to all tiers of Google’s AI Studio, the Gemini API, the Gemini App, and paid AI Pro and Ultra subscribers on the Antigravity development platform. The model can operate independently for tasks like writing and testing code or be paired with Carbon and Argon in complex workflows.
