AI Agents & Coding · Freemium
Gemini 4 Hydrogen
Gemini 4 Hydrogen is a Google DeepMind large language model in the Gemini 4 family, released to general availability on November 12, 2026. Its model ID is gemini-4-hydrogen. Google positions it as the fast, efficient member of the family for everyday coding and agentic workflows, alongside the larger frontier reasoning models Gemini 4 Carbon and Gemini 4 Argon, and as the recommended upgrade from Gemini 3.8 Flash and earlier Gemini 3 and 2.5 models.
What it is for: repository coding and debugging, interactive web development, tool-using agents, multimodal document work over PDFs, spreadsheets, images, audio and video, and parallel worker or verifier agents in multi-model setups. In Google Antigravity, a new Orchestrator option switches between Gemini 4 Carbon and Hydrogen in the same session, and the Antigravity agent in Gemini Managed Agents uses Hydrogen by default.
Key features: - A unified context window of 1,048,576 tokens shared by input, thinking and output, with long decode beyond the 65,536-token single-request cap through background execution or client-side continuation. - Four thinking levels: low, medium, high (the default) and a new xhigh level. The minimal level used on Gemini 3.8 Flash is rejected. - Implicit caching on prompts of at least 4,096 tokens with a 95% discount on cached input; explicit context caching is not supported. - Text, image, audio, video and document input, including Microsoft Office and OpenDocument files on supported APIs. - Available through the Gemini API (GenerateContent and the Interactions API), Google AI Studio, Android Studio, Gemini Enterprise Agent Platform on Google Cloud, Gemini for business, the Gemini app and Google Search. On Google Cloud it is listed as a short-term GA release with a minimum 45-day retirement notice.
Pricing type: freemium. The Gemini API free tier includes Hydrogen with rate limits. On the paid tier, as of November 13, 2026, standard pricing is $1.50 per million input tokens, $0.075 per million cached input tokens and $7.50 per million output tokens including thinking tokens, flat across the full context window. Batch and Flex are half that ($0.75 and $3.75), Priority is $2.70 and $13.50, and an invite-only Contributor tier matches the Batch rate in exchange for data use. Grounding with Google Search is billed separately per query.
Independent results: Artificial Analysis lists Hydrogen at 51 on its Intelligence Index v5.0.1 at the high setting and 54 at xhigh, with an estimated $1.87 and $3.16 per index task. Google's own launch figures include 57% on Terminal-Bench 4.1, up from 18.2% for Gemini 3.8 Flash.
Alternatives: Gemini 3.8 Flash and Gemini 3.5 Flash-Lite from Google; Claude Haiku 5.5 and Claude Sonnet 5.5 from Anthropic; GPT-6 Luna from OpenAI; and open-weight options such as DeepSeek V4.1 Flash and Kimi K3.1 Pro.
Official URL: https://ai.google.dev/gemini-api/docs/whats-new-gemini-4-hydrogen
