Gemini Agentic Video Cuts Analysis Costs by 66%, Token Use by 88%

rseroter · x · 2026-09-02

Google has introduced "Agentic Video" capabilities for Gemini models. According to the announcement, this feature reduces video analysis costs by up to 66% and token consumption by up to 88%, while improving accuracy by up to 7%.

The technology enables models to autonomously understand and analyze video content, supporting use cases such as monitoring and media analysis. The accompanying blog post details the underlying mechanics and specific application scenarios.

Related event: Gemini Launches Agentic Video Understanding, Cutting Token Use by 88%(16 posts)→

Original post →

More from Models

Models channel →