Gemini 4 Hydrogen

Gemini 4 Hydrogen is an AI model from Google that developers can plug into their own products, and the company positions it for speed and efficiency in everyday coding and work where the model carries out multi-step tasks on its own. Compared with the previous Flash model, Google says it reads long documents in one go through a single limit shared by input and output, and it can also keep writing very long outputs when a long-output feature is turned on. It accepts text, images, video, audio and PDF files, so it can be used to understand text and images (or video) together, but what it returns is text only. It also supports code execution, function calling, structured outputs and a mode where it thinks before answering, and you can try it in Google AI Studio. Before choosing it, note that the detailed API differences from the previous model are covered on a separate migration page, and that explicit caching is not supported.