Dr. DocBench - Text Edit Distance: leaderboard
Metric: Normalized Levenshtein edit distance of the extracted page text against the expert transcription, times 100 (0-100), 4,514 expert-annotated difficult pages sampled by parser failure from long books in 52 subject domains; each page parsed with the model's own document-to-markdown prompt (specialized parsers through their official APIs); lower is better. Source: arxiv.org. Saturation forecast: Around December 2026. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 19 |
| 2 | Kimi K2.5 | 19 |
| 3 | Gemini 3.1 Pro (Preview) | 22 |
| 4 | Claude Opus 4.6 | 22 |
| 5 | Qwen 3.5 122B A10B | 23 |
| 6 | Qwen 3.5 Plus | 25 |
| 7 | Qwen 3.5 Flash | 26 |
| 8 | GPT-4o | 37 |
| 9 | Nemotron Nano 12B v2 VL | 62 |
Interactive version: theaggregate.ai/benchmark?slug=dr-docbench-text-edit-distance · How It Works · Data refreshed daily, snapshot 2026-09-29.