PureDocBench - Real Degraded: leaderboard
Metric: Overall parsing score (0-100): mean of text accuracy (100 times one minus the text normalized edit distance), formula CDM and table TEDS, on the real degraded track (the same pages through four physical acquisition chains: phone photos of prints and of photocopies, screen photographs, and screenshots with social-media compression) of 1,475 source pages in 10 domains and 66 subcategories, rendered from LLM-written HTML/CSS with source-linked annotations for text, formulas, tables and reading order; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 58 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.6 Flash | 75.73 |
| 2 | Claude Opus 5 | 75.27 |
| 3 | GLM-5.3 Flash | 74.99 |
| 4 | Qwen 3.8 27B | 73.59 |
| 5 | Gemini 3.1 Pro (Preview) | 71.98 |
| 6 | Qwen 3.5 122B A10B | 69.85 |
| 7 | Kimi K2.6 (Non-reasoning) | 68.02 |
| 8 | Qwen 3.5 27B | 65.92 |
| 9 | Qwen 3.5 9B | 65.45 |
| 10 | Qwen 3.5 4B | 63.47 |
| 11 | Qwen 3.5 397B A17B | 62.7 |
| 12 | Qwen3.8-Flash-Next | 62.42 |
| 13 | GPT-5.6 Sol | 61.59 |
| 14 | Qwen 3.5 35B A3B | 60.59 |
| 15 | Qwen 3.6 35B A3B | 60.12 |
Interactive version: theaggregate.ai/benchmark?slug=puredocbench-real-degraded · How It Works · Data refreshed daily, snapshot 2026-10-07.