Google DeepMind launches agentic video understanding for Gemini models

Google DeepMind today announced agentic video understanding, a new capability for Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite that pairs the models’ core reasoning with native video tools to dynamically search, scan, and inspect video segments across visual frames, audio, and transcripts.

Across standard video analysis benchmarks, the feature reduces token consumption by up to 88% and analysis costs by up to 66%, while improving accuracy by up to 7%. These gains are most pronounced on long-form video, where static processing typically forces developers to trade off between cost and detail retention.

Gemini 3.7 Flash with agentic video understanding is positioned at the accuracy-to-cost Pareto frontier among tested models for video understanding, offering the best overall quality and cost efficiency among the three supported variants.

Unlike static processing, which ingests video at a fixed frame rate (commonly 1 FPS), agentic video understanding lets the model take an active, goal-directed role in determining what to watch, at what speed, and through which modality, fetching only the necessary moments and signals.

The feature is available today for video uploads and YouTube videos via the Gemini API in Google AI Studio and the Gemini Enterprise Agent Platform, enabling developers to activate it by setting the API configuration to 'agentic'.

Use cases highlighted include sub-second moment retrieval for precise automated editing, long-form needle-in-a-haystack search across multi-hour recordings, resampling for anomaly detection, and accurate counting of actions or objects over time.