AI BriefingIntroducing Pangram 4 (2 min read)
Key point
Pangram announced Pangram 4, its next-generation AI detection model with dramatically improved performance and accuracy.
Details
Pangram has released Pangram 4, its most powerful and accurate AI detection model to date. It boasts a scale more than 6 times larger than the previous model, with greatly improved ability to detect boundaries in mixed human-and-AI text and to identify AI-assisted text.
Key performance metrics are as follows:
- False Positive Rate: achieved 0.0041%, roughly 1 in 24,000 documents
- False Negative Rate: significantly reduced from 1.99% to 0.3396%
- Strong robustness: recorded a high detection rate of 98.83% even against results from 13 major commercial 'Humanizer' tools
- Versatility: maintains a false negative rate below 0.7% across all model families, keeping pace with the latest frontier models
In line with the model upgrade, the credit calculation method has also changed. Scanning has been broken down from the previous 1,000-word unit to 100-word units, improving efficiency for users, and image scanning has been added to all plans at no price increase.
API pricing has been set at $0.05 per 100 words, and the existing Pangram 3 will be supported until September 30, 2026.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
NSA Spending Billions to Test Frontier AI Models for Security Vulnerabilities
Open-source 'Proof of Capture' unveiled to counter Apple Reference Image, proving image authenticity via steganography
Local Tool 'Deckard' Released for Automatic AI Text Detection in Browsers, Performance Limitations vs. Pangram and Usage GuideVerify if a file was generated by Claude (2-minute read)