Google Gemini launches agentic video understanding, cutting analysis costs by up to 66%

Google announced agentic video understanding for its Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite models. This feature combines model reasoning with native video tools to dynamically search for key moments in videos, unlike the old method of reading frames at a fixed rate.

In tests, it cut analysis costs by up to 66%, reduced token use by 88%, and improved accuracy by 7%, especially for long videos. Developers no longer need to manually pick frames—Gemini decides what to watch, at what speed, and through which mode, like frames, audio, or text.

The feature is now live in the Gemini API on Google AI Studio and Gemini Enterprise Agent Platform. Developers can enable it by setting the processing mode to "agentic." Google plans to expand it to Gemini apps and support YouTube video analysis, and in coming months it will power the "Ask YouTube" feature with better answers from video content.