Gemini 3.8 Flash released, described as most intelligent Gemini model to date
ConcludedRan 5 daysin the edition since 29 Aug 20262 sourcesupdated 2 Sep 2026
Google DeepMind agentic video understanding for Gemini
The gist
Latest: Google DeepMind Launches Agentic Video Understanding for Gemini Flash
The approach cuts token usage and analysis costs substantially while improving benchmark accuracy, affecting developers building video analysis applications. The planned integration with YouTube's Ask YouTube feature would bring the capability to a large consumer audience.
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.
The full picture
Agentic video understanding launched September 1 for Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite. The system dynamically searches and inspects video segments across frames, audio, and transcripts rather than ingesting at a fixed frame rate. Google DeepMind states the feature reduces token consumption by up to 88% and analysis costs by up to 66%, while boosting accuracy by up to 7% with Gemini 3.7 Flash on standard benchmarks. Specific capabilities include sub-second moment retrieval, long-form needle-in-a-haystack search, anomaly detection, and accurate action and object counting. The feature is planned to power YouTube's 'Ask YouTube' function and roll out to Gemini app users across Flash and Flash-Lite models.
How it developed
Google DeepMind launched agentic video understanding for Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite on September 1, letting the models dynamically choose what to watch, at what speed, and through which modality rather than processing at a fixed frame rate.
The company says the feature drops token consumption by up to 88% and costs by up to 66%, while lifting accuracy by up to 7% on standard benchmarks, and places Gemini 3.7 Flash at the accuracy-to-cost Pareto frontier among tested models. New capabilities include sub-second moment retrieval, anomaly detection, and long-form needle-in-a-haystack search; Google DeepMind says the feature will power YouTube's 'Ask YouTube' and roll out to Gemini app users across Flash and Flash-Lite models.
First in the edition · 29 August 2026
Google DeepMind announces rollout of Gemini Omni 1.1 Flash
Gemini Omni 1.1 Flash released as a production-ready generative video model via the Gemini API
Sources
DeepMind Blog
Google DeepMind on X
Related
- Google DeepMind's Gemini 3.8 Flash TTS
- Google DeepMind releases Gemini Omni 1.1 Flash video model
- Google DeepMind Launches Gemini 3.8 Live with Real-Time Video Avatars
- Google DeepMind releases Gemini 3.7 Flash at half the price of 3.6 Flash, with large coding gains
BackgroundFrontier models, open models, and how they are compared
Get the daily email
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.