Google Releases Gemini 4 Hydrogen for Everyday Coding and Agent Workflows, With Flat Pricing Across a One-Million-Token Context Window

Google released Gemini 4 Hydrogen on November 12, positioning it as a fast model for everyday coding and agent execution. The published API price is $1.50 per million input tokens and $7.50 per million output tokens, unchanged across its one-million-token context window; the model offers low, medium, high and xhigh thinking levels. Hydrogen is adapted from Gemini 4 Carbon and is rolling out through the Gemini API, AI Studio, the Gemini app, Search, Gemini for business and Antigravity. Benchmark and efficiency figures are Google's own test results.

A Gemini 4 tier built for execution

Google released Gemini 4 Hydrogen on November 12, placing it alongside Gemini 4 Carbon and Argon for everyday coding, tool use and multi-step agent execution. The official model card says Hydrogen is adapted from Gemini 4 Carbon and can share up to one million tokens of input-output context across text, images, audio and video. Synchronous calls default to 64,000 output tokens; the one-million-token output option is for background long decoding.

Developers can use it through the Gemini API, AI Studio and Android Studio, while Google also lists the Gemini app, Search, Gemini for business, Google Sheets and Antigravity as distribution channels. Availability and controls are not identical across those products: free Gemini-app users receive Hydrogen through automatic model selection when appropriate, while Antigravity is listed for paid Google AI Pro and Ultra subscribers.

Long context stays on one price tier

Google lists API pricing of $1.50 per million input tokens and $7.50 per million output tokens, and says that rate is flat across the one-million-token context window. Hydrogen has four thinking levels—low, medium, high (the default) and xhigh—giving teams a choice between lower-latency batch work and more extended reasoning. For teams moving from Gemini 3.8 Flash, that price and context policy is more actionable than an isolated benchmark: they should test their own prompts, tool failures and retry costs before choosing a default setting.

Vendor figures are not an integration test

Google's post says Hydrogen scores 51 on the Artificial Analysis Intelligence Index at the default high setting and 54 at xhigh; its model card lists 57.0% on Terminal-Bench 4.1. These are vendor-reported or vendor-cited evaluation results, not evidence of behavior in a particular codebase, business dataset or tool-permission setup. The model card also notes possible hallucinations, occasional slowness or timeouts, and higher token use at higher thinking settings. Google also introduced Orchestrator in Antigravity to switch dynamically between Carbon and Hydrogen; its quota and speed claims likewise remain vendor test figures.

via: Google Blog: Introducing Gemini 4 Hydrogen for Agentic Workflows, Google DeepMind: Gemini 4 Hydrogen Model Card