OCRTurk - Tables: leaderboard
Metric: Tree-edit-distance similarity (TEDS, 0-1) between the output and reference HTML table trees over the benchmark's 130 tables; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek-OCR | 0.87 | |
| 2 | OCRTurk PaddleOCR-VL (checkpoint unspecified) | 0.87 | |
| 3 | Docling | 0.86 | |
| 4 | NVIDIA-Nemotron-Parse-v1.1 | 0.86 | |
| 5 | HunyuanOCR | 0.85 | |
| 6 | olmOCR-2 | 0.85 | |
| 7 | Nanonets-OCR2-3B | 0.8 |
Interactive version: theaggregate.ai/benchmark?slug=ocrturk-tables · How It Works · Data refreshed daily, snapshot 2026-10-11.