Pangram says its new AI text detector makes only one mistake per 24,000 documents
By Matthias Bastian
AI 摘要
Pangram 4 detects 99.66 percent of AI-generated text with just one false positive per 24,000 documents, the company claims. The model also resists "humanizer" tools that disguise AI writing as human. API prices go up two- to tenfold. The article Pangram says its new AI text detector makes only on
原文正文
Pangram says its new AI text detector makes only one mistake per 24,000 documents
Pangram has introduced Pangram 4, a new model for detecting AI-generated text. According to the company’s own benchmarks, it correctly identifies 99.66 percent of all AI texts and falsely flags human-written texts as AI-generated in only 0.0041 percent of cases—which corresponds to about one error per 24,000 documents. According to the manufacturer, Pangram 4 is six times larger than its predecessor and produces 14 times fewer false positives and six times fewer missed AI texts than Pangram 3.
Pangram 4 is also said to be able to distinguish between text that has been lightly polished by AI and text generated entirely by AI, and to be robust against tools designed to make AI-generated text sound more human: for 13 common "humanizers," the model detects the AI component in 98.83 percent of cases.
The API costs $0.05 per 100 words, which, depending on the document length, corresponds to a price increase of two to ten times. In addition, all plans can now scan images at no additional cost. Pangram 3 will continue to be supported for two more months and will be shut down on September 30, 2026.
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.
Subscribe now