Introducing Agentic Video Understanding with Gemini
Google, Tuesday, September 1st, 2026
Google's agentic video feature cuts token use up to 88% and costs up to 66% while improving quality up to 7%.
Google launched agentic video understanding across its latest Gemini models, including 3.7 Flash and 3.6 Flash. Instead of ingesting an entire video as tokens, the agentic approach lets the model navigate and sample the footage as needed to answer a question.
Google reports that this cuts token consumption by up to 88 percent, reduces costs by up to 66 percent, and improves quality by up to 7 percent.
The feature is aimed at developers building video analysis applications where full-video ingestion has been prohibitively expensive.