Pangram detects Claude Opus 5.5 with 99.7% accuracy; "in short" is the new em dash
SuB8u · x · 2026-09-26
AI-detection company Pangram ran 1,000 Chatbot Arena prompts through Claude Opus 5.5: Pangram 4 hit 99.7% accuracy, classifying 997 outputs as fully AI, 2 as mixed, and just 1 as human — that lone miss being a Club Penguin parody poem in the style of Ozymandias.
False negatives clustered in literary text (a Jane Austen chapter explanation, a Thucydides-style speech). The piece also notes Opus 5.5's new writing fingerprint — heavy use of "in short," this generation's em dash — and judges it the best-writing Claude yet.
More from Models
- OpenRouter launches Jev Router, a cache-aware router that scores prompt difficulty to pick models — kalyan_kpl · 2026-09-26
- System 1 models return typed answers in one encoder pass: 45ms per query on a laptop CPU, but often confidently wrong — ghumare64 · 2026-09-26
- Open System 1 model runs at 45ms per query on laptop CPU, 70% on support tickets — ghumare64 · 2026-09-26
- Hands-on with Laya, an open System 1 model: one encoder pass returns typed probabilities in milliseconds — ghumare64 · 2026-09-26
- Stanford professor: LLMs are great brainstorm partners but confidently spew plausible nonsense — anshulkundaje · 2026-09-26
- Researcher says Claude solved his months-old multimode spectrum problem in hours — jwt0625 · 2026-09-26