Agentic Video Understanding in Gemini

the percentage cuts in tokens and costs caught my attention for real. And at that speed?

Agentic video understanding is a new processing mode Google just shipped across Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite. Shoutout to @rkdoshi and team! :) Instead of scanning footage at a fixed frame rate, the model actively decides what to watch, at what speed, and through which modality (frames, audio, transcript), fetching only the segments it needs via an internal agentic loop.What makes it different: it's not a separate product, it's a processing mode toggle inside models you likely already use, with standard token pricing and no added fee.Key features:Dynamic scanning: model chooses what to inspect and at what FPSUp to 88% fewer tokens, up to 66% lower cost, up to 7% better accuracySub-second moment retrieval for precise auto-editingNeedle-in-haystack search across multi-hour videoAnomaly detection via variable FPS resamplingAccurate counting of repeated actions/objectsWho it's for: developers and teams building video search, editing, moderation, or analysis tools on long-form content. Try at Google AI StudioP.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends