Pangram has launched Pangram 4, a new AI text detection model designed to identify AI-generated content with high accuracy. The model is six times larger than its predecessor and has 14 times fewer false positives, according to the company. In internal testing, Pangram 4 correctly identifies 99.66 percent of AI-generated text and flags human writing just 0.0041 percent of the time, which translates to roughly one error per 24,000 documents. The model also distinguishes between AI-assisted text and mixed human-AI writing, as well as between lightly AI-polished text and fully AI-generated content. It effectively detects AI components in 98.83 percent of cases across 13 common humanizer tools, the company said. The API for Pangram 4 costs $0.05 per 100 words, with pricing varying depending on document length. All plans now include image scanning at no extra cost, and Pangram 3 remains available until September 30, 2026.
Pangram's business has seen significant growth, with annual revenue increasing 35 times year over year. Monthly users rose from 2,700 in June 2025 to 120,000 in June 2026, according to the New York Times. The company's expansion reflects growing demand for tools that can detect AI-generated content in an era of increasing AI adoption.
Pangram 4 is part of the company's ongoing efforts to improve the accuracy and reliability of its AI detection capabilities. The model's performance is based on internal benchmarks, and the company has not provided details on external testing or comparisons to other models.
Source: thedecoder
