Pangram launches AI image and video detectors with 99.96% accuracy
The flagship model, Pangram Image 1.0, achieves 99.96% accuracy on a test set of 1,200 synthetic images created by 12 frontier AI generators. It includes a heatmap feature to pinpoint AI content within images and claims robustness against edits, compression, and low resolutions (now supporting 256x256 pixels with reduced confidence disclaimers below 512x512). The smaller Flash 1.0 model excels…
Key points
- Pangram Image 1.0 achieves **99.96% accuracy** detecting AI-generated images in a test of 1,200 synthetic images
- Flash 1.0 scores **98.7% on DF26**, a **29-point improvement** over existing video detectors
- Both models now include **Chrome extensions** and **API support**, plus native video upload for frame-by-frame analysis
The company tested its tools against academic benchmarks like X-AIGD, Mirage-Test, and AIGVDBench, comparing performance against commercial and open-source detectors. Pangram’s models also support native video uploads for frame-by-frame analysis. The release follows a research preview earlier this year, with improvements driven by community feedback. Isaac, a research engineer at Pangram (formerly Samsung AI Center NYC), led the development.
Pangram releases Pangram Image 1.0, which it says is state-of-the-art in AI image and AI video detection, and the smaller Image Flash 1.0 for AI video detection
pangram.com · 4 November 2026
Loading the full article…
This text was published by pangram.com and written by Isaac Kasahara. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
2sources- Hacker News discussion · 5 pointsnews.ycombinator.com
- Pangram Launches Two Models for AI Image and Video DetectionPress · Unite.AI ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Ricursive Intelligence releases Ripple-65B Preview, a 65B model with 1M context · 1 src
- Kimi-K3.1-Pro is now open-weights · 1 src
- Cognition says SWE-2 Flash runs three times faster and costs one-fifth · 1 src
- A new way of measuring AI models suggests China's threat to OpenAI and Anthropic may have been overblown · 4 src
- structRFM model trains on 21 million RNA pairs · 1 src
Comments
via GitHub Discussions