Superpower Daily
Altruist Puts Six-Part Financial Planning in Hazel, With Separa...Meta Introduces Muse Voice Transcribe for Live Speaker Labels i...Anthropic’s EFS Puts Misuse-Monitoring Data in Customer Clouds,...Grok 4.6 Clears 50% on Biosecurity Refusal and Routine Biology...HUMAIN Picks Compal for Cloud-to-Edge AI Build, With Nvidia and...t54 Puts a Hard Trust Gate on Bedrock Agent Payments, Citing 20...Altruist Puts Six-Part Financial Planning in Hazel, With Separa...Meta Introduces Muse Voice Transcribe for Live Speaker Labels i...Anthropic’s EFS Puts Misuse-Monitoring Data in Customer Clouds,...Grok 4.6 Clears 50% on Biosecurity Refusal and Routine Biology...HUMAIN Picks Compal for Cloud-to-Edge AI Build, With Nvidia and...t54 Puts a Hard Trust Gate on Bedrock Agent Payments, Citing 20...
Modelspublished

Google Adds Agentic Video Understanding to Gemini, Claiming 88% Lower Token Use

The API feature shifts Gemini from fixed-rate video sampling to targeted inspection of frames, audio and transcripts. Its value for long recordings will depend on whether Google’s reported efficiency and accuracy gains carry into production workloads.

By 3 min read
Google Adds Agentic Video Understanding to Gemini, Claiming 88% Lower Token Use
Google Adds Agentic Video Understanding to Gemini, Claiming 88% Lower Token Use
Sign in to save and react

Listen to this story

The audio brief

About 1:20
0:001:20
Read transcript
Google is giving Gemini a new way to watch video: instead of sampling frames at a fixed rate, it can decide which moments to inspect, and whether frames, audio, or transcripts contain the answer. The feature, called agentic video understanding, is available now through the Gemini API for uploaded and YouTube videos. Google says it cut token use by as much as 88 percent, reduced analysis cost by up to 66 percent, and improved accuracy by up to 7 percent on standard video-analysis benchmarks. Those are company-reported results, so they indicate tested performance—not a guarantee for every task. The supported models are Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite. Developers turn the setting on in Google AI Studio or the Gemini Enterprise Agent Platform, with standard Gemini API pricing and no separate feature fee. The practical change is selective resampling: the system can revisit a short stretch at a higher rate when it needs to catch a fast movement, a subtle change, or an exact cut. Google highlights sub-second moment retrieval, searching multi-hour recordings, anomaly detection, and counting actions or objects. The feature is also planned for the Gemini app and for YouTube’s Ask YouTube experience. The key question is whether Google’s efficiency and accuracy gains hold up across real production videos, not just benchmarks.

Story brief

3 key points

Google is shifting Gemini video analysis from uniform frame sampling to an internal agentic loop that chooses when to inspect and whether to use frames, audio, or transcripts. The mode is available now through the Gemini API for uploaded and YouTube videos, with no separate feature fee. Google reports up to 88% lower token use, 66% lower cost, and 7% higher accuracy on standard benchmarks—results that may vary by...

  1. 01

    Supports Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite.

  2. 02

    Developers enable agentic processing in Google AI Studio or Gemini Enterprise Agent Platform.

  3. 03

    Selective resampling targets sub-second retrieval, long-video search, anomalies, and action or object counting.

Google has launched agentic video understanding for Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite. The feature lets the models search selected parts of a video, using images, audio and transcripts instead of processing the full recording at a fixed frame rate. Google says this can cut token use by up to 88% on standard video-analysis benchmarks.

The capability is available now for uploaded videos and YouTube videos through the Gemini API in Google AI Studio and the Gemini Enterprise Agent Platform. Developers enable it by setting video processing to “agentic”; Google says it uses standard Gemini API token pricing and carries no separate feature fee.

From uniform sampling to targeted inspection

Static video processing ingests media at a set frames-per-second rate. Google says agentic video understanding instead combines Gemini’s reasoning with native video tools, allowing the model to decide what moment to load, how quickly to inspect it, and whether frames, audio, or a transcript offer the relevant signal.

The change affects both workflow and model behavior. Developers could previously build targeted inspection themselves, Google says. The new mode moves that selection into an internal agentic loop for long videos, where fixed-rate processing can trade higher token use for missed details.

Google’s reported benchmark results
Up to 88% lowerToken consumption

On standard video-analysis benchmarks, Google reports up to 88% lower token consumption.

Up to 66% lowerAnalysis cost

The company also reports analysis costs up to 66% lower on those benchmarks.

Up to 7% higherAccuracy

Accuracy was up to 7% higher, according to Google’s benchmark results.

The jobs that benefit from a second look

Google identifies four intended uses: retrieving a precise moment in under a second, searching long recordings for hard-to-find details, detecting anomalies, and counting actions or objects. Each relies on selective resampling. The system can revisit a short window at a higher rate when fast motion or a subtle change requires closer inspection.

  • Sub-second moment retrieval is intended to find brief state changes and cut boundaries that a one-frame-per-second pass can miss.
  • Long-form search is intended to answer complex questions across multi-hour video without processing the recording uniformly.
  • Anomaly detection and action or object counting can use closer inspection of relevant time windows.

A configuration choice, not a new price tier

The product claim is not that every video needs deeper analysis. It is that the model can fetch only the moments and modalities needed for a particular request. Google’s cost, token, and accuracy figures are company-reported benchmark results, so they describe tested performance rather than a guarantee for every video or task.

A route from developer API to consumer surfaces

Google plans to roll the feature out to Gemini app users on Flash and Flash-Lite models. It also says agentic video understanding will power YouTube’s Ask YouTube feature on video watch pages in the coming months, with answers intended to be grounded in video visuals.

Editorial analysis

Our Read

Google’s strategic move is to make selective video inspection a built-in model behavior rather than a workflow developers must assemble themselves. The promised gains are most consequential for long recordings, where fixed sampling can either miss fast events or consume far more tokens. Google’s planned Ask YouTube rollout will be a visible test of that premise. The next meaningful evidence will be production results beyond Google’s benchmarks: whether long lectures, multi-hour recordings, and rapid-motion footage retain the claimed cost and quality advantages when the model chooses what to revisit.

Sources

  1. blog.googleIntroducing agentic video understanding with Gemini

Related coverage

Continue reading