Date: September 1, 2026 Main source: Google Link: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-agentic-video-in-gemini/
Summary
Google launched agentic video understanding for Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite. The system lets the model dynamically examine only the video segments it needs to answer, instead of processing all the content the same way. Google says the approach can cut token use by up to 88%, lower costs by up to 66% and improve quality by up to 7%. The feature is already available through Google AI Studio and the Gemini Enterprise Agent Platform.
Why it matters
Video analysis has been one of the most expensive multimodal workloads because of the amount of information that has to be processed. A reduction this size opens use cases that used to be hard to justify economically, such as reviewing long recordings, inspections, training, support, security and media content. The pattern also points to a shift from models that consume the whole context to agents that decide what information they need to inspect.
Practical implications
- Businesses that handle many hours of video should recalculate the cost of automating analysis and classification.
- Agencies can test this capability for content reuse, extracting key moments and visual audits.
- Teams should measure cost per completed task, not just price per token, because the model now decides how much video to inspect.