Gemini 4 Hydrogen

Gemini 4 Hydrogen is Google’s fast, cost-conscious model for everyday coding, tool-using agents, and multimodal workflows. Available on free and paid tiers through AI Studio and the Gemini API, it has a unified context window of 1,048,576 tokens, combining input, reasoning, and output. Google says it produces cleaner code and makes more than five times fewer fabricated claims than Gemini 3.8 Flash by flagging missing information rather than guessing. It can run read and search tools in parallel, recover more effectively from failed tool calls, and interpret tables, charts, PDFs, images, audio, and video, including by inspecting specific video frames. It is also the default model for Managed Agents.

Hydrogen offers four thinking levels: low for throughput and low latency, medium for responsive interactive tasks, high as the general-purpose default, and xhigh for difficult multistep problems at greater cost and latency. Paid pricing is $1.50 per million input tokens and $7.50 per million output tokens. Synchronous responses can reach 65,536 tokens; longer generations, up to the context limit, require background execution or SDK continuation in 64K-token steps. The guidance positions Hydrogen as an upgrade from Gemini 3.8 Flash and earlier models, while Carbon suits the hardest reasoning and verification tasks, Argon handles advanced long-horizon work with lower token use than Carbon, and Flash-Lite targets high-volume, very low-latency tasks. Carbon and Hydrogen can also divide planning and implementation across agents.