Pangram released the AI text detector Pangram 4 and secured an investment of 9 million dollars
Pangram released the AI text detector Pangram 4 (according to the company, accuracy of 99.66 %, an error rate of approximately 1 per 24 000 documents) and secured an investment of 9 million dollars led by Menlo Ventures. The API price increased 2–10×, and Pangram 3 ends on 30. 9. 2026.
Pangram released a new version of its AI-generated text detector called Pangram 4 and, according to TechCrunch, also announced an investment of 9 million dollars in a round led by Menlo Ventures with participation from Haystack, ScOp, Script Capital and Cadenza. The company also launched a new model for detecting AI-generated images, Pangram Image, which is currently available only as a research preview; the company says it plans a broader release in the coming weeks.
According to Pangram, the Pangram 4 model correctly identifies 99.66 % of AI-generated texts and incorrectly flags human-written text as AI-generated in only 0.0041 % of cases, equivalent to roughly one error per 24 000 documents. According to the company, the model is six times larger than its predecessor Pangram 3, produces fourteen times fewer false positives and misses six times fewer AI texts. It can also distinguish text only lightly edited by AI from text generated entirely by AI and, according to the company, is resistant to text "humanization" tools — it detects AI involvement in 98.83 % of cases across 13 common "humanizers". However, TechCrunch quotes co-founder Max Spero giving a different error rate — approximately one incorrectly flagged document per 10 000 — so the sources disagree on the exact figure.
The API now costs 0.05 dollars per 100 words, which, according to the company, is roughly a twofold to tenfold price increase over the previous version, depending on document length; image scanning is now included in all plans at no extra charge. Support for Pangram 3 will continue for another two months and end on 30 September 2026. According to TechCrunch, the company was founded two years ago by Stanford graduates Max Spero and Bradley Emi, and its API customers include Substack, Quora, schools and universities, publishers and recruiters; competitors include Winston AI, Originality.ai, Copyleaks and GPTZero.
Testing by a TechCrunch reporter showed that the tool reliably recognized texts generated by ChatGPT and Claude and resisted attempts to "humanize" the text, but also occasionally flagged sentences that had been entirely rewritten by a human as AI-generated. Details on the testing and the Pangram Image model can be found in the source article.
Why it matters
AI text detection is becoming part of mainstream platforms and institutions (Substack, schools, publishers, recruitment), so the accuracy and price of these tools directly affect whose texts will be flagged as AI-generated and at what cost. Sources disagree on the reported error rate, which matters when assessing the risk that human-written text will be incorrectly "detected".
What was added since the original report
Verified updates
-
Pangram 4 detects 99.66 % of AI texts with an error rate of 0.0041 % (1 error per ~24 000 documents); The model is 6× larger than Pangram 3 with 14× fewer false alarms; The API costs $0.05 per 100 words; a twofold to tenfold price increase over Pangram 3; Pangram 3 will be shut down on 30 September 2026; The model detects AI content in 98.83 % of cases and is resistant to 13 common humanizers; image scanning at no additional cost
- Pangram 4 detects 99.66 % of AI texts with an error rate of 0.0041 % (1 error per ~24 000 documents)
- The model is 6× larger than Pangram 3 with 14× fewer false alarms
- The API costs $0.05 per 100 words; a twofold to tenfold price increase over Pangram 3
- Pangram 3 will be shut down on 30 September 2026
- The model detects AI content in 98.83 % of cases and is resistant to 13 common humanizers; image scanning at no additional cost
Two audiences, two different impacts
What this means
For individuals
People whose texts (school assignments, articles, applications) are checked by similar tools may encounter stricter detection as well as the risk that entirely human-written text will be incorrectly flagged as AI-generated.
For a business
Organizations using the API from Pangram (according to TechCrunch, these include Substack, schools, publishers and recruiters) face a twofold to tenfold increase in the price of the service and the need to migrate from Pangram 3 to Pangram 4 by the end of September 2026.
Risks and complianceCheck the original
Event sources
confirmed by 2 independent sources · 2 publishers, 2 independent. We count feeds from the same owner only once.