⚡ Uncle Cat AI Radar
ModelsAgentsIndustry

Google rolls out Gemini 4 Hydrogen for faster agent work

Google’s new Gemini 4 Hydrogen model targets rapid coding and agent execution, combining lower cost, broad access and multi-model orchestration.

A faster Gemini tier

Google has introduced Gemini 4 Hydrogen, a model positioned between everyday assistant use and the company’s heavier reasoning systems, Gemini 4 Carbon and Argon. Hydrogen is available immediately across the Gemini app, Google Search, AI Studio, the Gemini API, enterprise products and selected developer tools.

Google says Hydrogen surpasses Argon on the Artificial Analysis Intelligence Index at its default high-thinking setting, scoring 51 against 50, while completing many tasks faster. Its pricing is set at $1.50 per million input tokens and $7.50 per million output tokens across a one-million-token context window.

Designed as an execution layer

The more consequential announcement is Google’s Orchestrator in Antigravity. It can shift work between models within one session: a stronger model can plan or coordinate, while Hydrogen handles coding, testing and verification. Google reports that this setup reached near-Carbon success rates on Terminal-Bench 4.0 while using 24% less quota and finishing 19% faster.

Hydrogen is also being marketed as a practical agent component rather than only a chat model. Google highlights tool use, computer use, coding and resistance to indirect prompt injection, claiming more than twice the resilience of an earlier baseline on its Gray Swan benchmark.

Why it matters

This release broadens Gemini 4 from a frontier-model family into a deployment stack. Google is putting a relatively capable model inside Search and consumer products while making it cheap enough to serve as the worker model in larger agent systems. The unresolved question is whether Google’s internal benchmarks translate into reliable production behavior when agents are allowed to act across real software, files and business tools.

Uncle Cat take: Hydrogen’s important number is the 24% quota reduction in Orchestrator, because Google is selling coordination economics, not merely another leaderboard score.

Sources