PureDocBench - Clean: 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 clean track (pages rendered straight from their HTML/CSS source) 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 July 2028. 58 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GLM-5.3 Flash | 83.04 |
| 2 | Claude Opus 5 | 82.99 |
| 3 | Qwen 3.8 27B | 80.45 |
| 4 | Gemini 3.6 Flash | 79.97 |
| 5 | Qwen 3.5 122B A10B | 76.14 |
| 6 | GPT-5.6 Sol | 74.15 |
| 7 | Qwen 3.5 9B | 73.87 |
| 8 | Qwen 3.5 4B | 73.45 |
| 9 | Kimi K2.6 (Non-reasoning) | 72.32 |
| 10 | Qwen3.8-Flash-Next | 72.23 |
| 11 | Qwen 3.6 35B A3B | 72.14 |
| 12 | Qwen 3.5 27B | 72.07 |
| 13 | Qwen 3.6 27B | 70.2 |
| 14 | Gemini 3.1 Pro (Preview) | 70.04 |
| 15 | Qwen 3.5 397B A17B | 69.12 |
Interactive version: theaggregate.ai/benchmark?slug=puredocbench-clean · How It Works · Data refreshed daily, snapshot 2026-10-07.