Google has launched agentic video understanding for its latest Gemini models, including Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite. This new capability allows the model to take an active, goal-directed role in determining what to watch, at what speed, and through which modality—frames, audio, or transcript—fetching only the necessary signals for a given task.
From Static to Agentic Processing
Unlike traditional "static" processing, where a model ingests video at a fixed frames-per-second (FPS) rate, agentic video understanding pairs core reasoning with native video tools to dynamically search and scan target segments. According to Google, this approach reduces token consumption by up to 88% and costs by up to 66%, while improving overall accuracy by up to 7%.
Enhanced Capabilities for Long-Form Content
The efficiency gains are particularly significant for long-form video, such as multi-hour recordings or lectures. New capabilities enabled by this feature include:
- Sub-second moment retrieval: Pinpointing split-second state changes for precise automated editing.
- Anomaly detection: Resampling specific time windows at higher FPS to inspect rapid motions.
- Precise counting: Tracking repeated physical movements or distinct objects over time.
- Needle-in-a-haystack search: Answering complex queries across long videos without consuming millions of tokens.
The feature is available today via the Gemini API in Google AI Studio and the Gemini Enterprise Agent Platform. Google plans to roll out these improvements to the Gemini app and the "Ask YouTube" feature in the coming months.
