PureDocBench - Real Degraded - Table TEDS: leaderboard
Metric: Table Tree-Edit-Distance-based Similarity (0-100) between the predicted and reference HTML tables, 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 December 2027. 58 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 | 84.01 |
| 2 | GLM-5.3 Flash | 82.45 |
| 3 | Gemini 3.6 Flash | 81.96 |
| 4 | Qwen 3.8 27B | 79.63 |
| 5 | Gemini 3.1 Pro (Preview) | 77.26 |
| 6 | Qwen 3.5 122B A10B | 75.44 |
| 7 | Kimi K2.6 (Non-reasoning) | 75.14 |
| 8 | GPT-5.6 Sol | 69.75 |
| 9 | Qwen 3.5 9B | 68.59 |
| 10 | Qwen 3.5 4B | 67.17 |
| 11 | Qwen3.8-Flash-Next | 66.85 |
| 12 | Qwen 3.6 27B | 66.34 |
| 13 | Qwen 3.5 27B | 64.82 |
| 14 | Qwen 3.6 35B A3B | 63.27 |
| 15 | Qwen 3.5 2B | 60.79 |
Interactive version: theaggregate.ai/benchmark?slug=puredocbench-real-degraded-table-teds · How It Works · Data refreshed daily, snapshot 2026-10-07.