LATEST

← Back to GPN · All articles

Google DeepMind Introduces Gemini 4 Hydrogen for Fast, Everyday AI Work

Google DeepMind Introduces Gemini 4 Hydrogen for Fast, Everyday AI Work

OFFICIAL Written by GPN · Nov 13, 2026, 5:23 AM ET · · AI

Google DeepMind has introduced Gemini 4 Hydrogen, a new model built for speed and efficiency on everyday work. It sits alongside the company's frontier models, Gemini 4 Carbon and Argon, which handle deep, long-horizon reasoning, while Hydrogen is aimed at fast execution across coding and agentic workflows. It is rolling out globally starting today across the Gemini API, Google AI Studio, the Gemini app, Google Search, and Google Antigravity.

On benchmarks, Hydrogen edges past Argon on the Artificial Analysis Intelligence Index at its default high thinking level, scoring 51 against Argon's 50, and reaching 54 at its highest xhigh setting. Compared with the model it succeeds, Gemini 3.8 Flash, DeepMind says Hydrogen finishes tasks 2.5 times faster at its high default setting while more than tripling its Terminal-Bench 4.1 score, from 18.2 percent to 57 percent, and cutting fabricated claims by more than five times.

Hydrogen comes with four thinking levels - low, medium, high, and a new xhigh tier - so users can trade reasoning depth for speed and cost. Pricing is flat across its entire 1 million-token context window at $1.50 per million input tokens and $7.50 per million output tokens.

DeepMind is also introducing Orchestrator in Google Antigravity, which switches between models inside a single session instead of running one model start to finish. On Terminal-Bench 4.0, Orchestrator reached near-Carbon task success rates while using 24 percent less quota and finishing 19 percent faster. Carbon can also act as a coordinator while Hydrogen powers worker and verifier agents in parallel.

For everyday use, Hydrogen is available in the Gemini app and in Search for all users, in the Gemini API and AI Studio for developers, in Android Studio, and in Gemini for business and Google Sheets for eligible subscriptions. Paid Google AI Pro and Ultra subscribers can also try it in Antigravity.

On safety, DeepMind says Hydrogen reduces overreach in agentic tasks by up to 81 percent and under-service by up to 79 percent, cuts misaligned actions and task shortcutting across tool-use and coding evaluations by 12 times, and reaches a 96.7 percent honesty rate on its internal agent honesty evaluation. It is also more than 2.3 times more resilient to indirect prompt injection attacks on the Gray Swan IPI benchmark.

Facts gathered from

More on this