Google Gemini's new agent-based video analysis cuts token usage by up to 88 percent

AITopTools Editorial TeamSeptember 2, 2026

What changed

Google is adding agent-based video analysis to Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite. Instead of reviewing every frame at one fixed rate, the systems choose which parts of a video to examine and how much detail to use. Google says this can reduce usage by up to 88% and improve accuracy, particularly for videos lasting several hours.

What this means for you

Developers using these Gemini models may be able to analyze long videos more efficiently and accurately, potentially lowering usage costs. The metadata says the feature is being added, but does not specify when it will be available or its pricing.

Related AI tools

Explore directory listings connected to the products, companies, and workflows in this story.

LivePhoto.video Aylien Text Analysis API Video Face Swap Video Transcriber

Related AI news

New AI featuresSep 4, 2026

Google DeepMind’s WeatherNext 3 Trains on Weather Station Observations to Deliver 5 km Global Forecasts, Refreshed Every Hour

People using these Google services can get more frequently updated, detailed weather information without opening a separate forecasting tool. The metadata does not specify regional coverage limits, pricing, or when the update will reach every user.

MarkTechPostSee why it matters

New AI featuresSep 3, 2026

OpenAI Releases GPT-6 Astra: A 1.05M-Context Computer-Use Model Gated Behind a ‘Critical’ Cyber Threshold

Astra may help eligible users automate computer-based tasks, but access and permitted uses are limited by OpenAI’s “Critical” cybersecurity classification. Its listed price is $10 for input and $50 for output per million text units processed.

MarkTechPostSee why it matters

AI toolsSep 3, 2026

Anthropic Released Claude Commerce Agents: An Apache-2.0 Blueprint for Shopping and Merchant Agents Across Retail, Travel, Telecom and Entertainment

Teams can use the published code as a starting point instead of rebuilding this basic infrastructure themselves. It is available under a permissive open-source license, but the feed does not specify how ready it is for production use or what it costs to operate.

MarkTechPostSee why it matters