How much does agentic video processing cost in the Gemini API?

Agentic video processing uses standard Gemini API token pricing with no additional feature fee. You pay for the tokens the model actually loads, and because it skips the parts of the video it does not need, Google reports up to 88% fewer tokens and up to 66% lower cost than static, fixed-rate processing on long-form content.

Which Gemini models and video sources support agentic processing?

Google released agentic video understanding for Gemini 3.7 Flash, Gemini 3.6 Flash and Gemini 3.5 Flash-Lite, across both the Interactions and GenerateContent APIs. It works on video you upload and on YouTube links, through Google AI Studio and the Gemini Enterprise Agent Platform. Static, fixed-frame-rate processing remains the default when you do not set the flag.

What can I ask a long video now that was impractical before?

Google names sub-second moment retrieval, long-form needle-in-a-haystack search, anomaly detection, and accurate counting of actions and objects as the use cases agentic video opens up. These all involve scanning a long recording for something specific, which was expensive when every minute of footage cost tokens whether or not it answered the question.

Is agentic video available in the Gemini app or on YouTube?

Not yet. Agentic video understanding is available today only through the Gemini API. Google says it is rolling out to Gemini app users soon, and coming to YouTube's "Ask YouTube" feature in the coming months. Developers who want it now have to call the API and set processing to agentic in the request config.

Google · 2026-09-01 · major

Agentic video in Gemini — the model loads only the clips it needs

Gemini's API now offers agentic video processing on Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite. The model walks the video and loads only the frames, transcript or audio it needs, using up to 88% fewer tokens than static processing.

Google blog header for agentic video understanding in Gemini

Gemini can now walk a video on its own, pulling only the segments a question needs instead of every frame at a fixed rate.

Quick facts

Maker	Google
Released	1 September 2026
Models	Gemini 3.7 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite
How to turn it on	Set "processing": "agentic" in the API config
Token use	Up to 88% fewer than static processing
Cost	Up to 66% lower, no extra feature fee
Where	Gemini API via Google AI Studio and the Gemini Enterprise Agent Platform

What is it?

Agentic video processing is a new mode in the Gemini API that lets the model choose which parts of a video to look at. Google turned it on for Gemini 3.7 Flash, Gemini 3.6 Flash and Gemini 3.5 Flash-Lite on 1 September 2026. Developers switch it on by setting processing to "agentic" in the request config, and it works on uploaded video and on YouTube links.

How does it work?

Static processing samples a video at a fixed frame rate and pushes all of it into the context. In agentic mode the model issues requests as it reasons instead: the response stream carries a processing_call step when it asks for a segment, transcript or audio track, and a matching processing_result step with what came back, linked by call_id. Google's docs report up to 88% fewer tokens and about 7% higher quality on long-form content.

Why does it matter?

Long video used to be priced by its length rather than by the question, so an hour of footage cost the same whether the answer sat at minute 3 or minute 53. Google puts the cost drop at up to 66%, which changes what is worth building: searching a long recording for one moment, spotting an anomaly in security footage, or counting actions across a match. There is no extra feature fee — you pay the standard token rate for what the model loads.

Who is it for?

developers building video search and analysis

Frequently asked questions

How much does agentic video processing cost in the Gemini API?: Agentic video processing uses standard Gemini API token pricing with no additional feature fee. You pay for the tokens the model actually loads, and because it skips the parts of the video it does not need, Google reports up to 88% fewer tokens and up to 66% lower cost than static, fixed-rate processing on long-form content.
Which Gemini models and video sources support agentic processing?: Google released agentic video understanding for Gemini 3.7 Flash, Gemini 3.6 Flash and Gemini 3.5 Flash-Lite, across both the Interactions and GenerateContent APIs. It works on video you upload and on YouTube links, through Google AI Studio and the Gemini Enterprise Agent Platform. Static, fixed-frame-rate processing remains the default when you do not set the flag.
What can I ask a long video now that was impractical before?: Google names sub-second moment retrieval, long-form needle-in-a-haystack search, anomaly detection, and accurate counting of actions and objects as the use cases agentic video opens up. These all involve scanning a long recording for something specific, which was expensive when every minute of footage cost tokens whether or not it answered the question.
Is agentic video available in the Gemini app or on YouTube?: Not yet. Agentic video understanding is available today only through the Gemini API. Google says it is rolling out to Gemini app users soon, and coming to YouTube's "Ask YouTube" feature in the coming months. Developers who want it now have to call the API and set processing to agentic in the request config.

Try it

Set "processing": "agentic" in your Gemini API request config