Page-level OCR on OmniDocBench English DODO evaluation setting
3.2Normalized Edit Distancedots.ocr
Evaluation Results
| Method | Links | |
|---|---|---|
| dots.ocrSize=3B, Category=Specialized OCR2026.05 | 3.2 | |
| MinerU 2.0 VLMSize=0.9B, Category=Specialized OCR2026.05 | 4.5 | |
| DeepSeek-OCRSize=3.4B, Category=Specialized OCR2026.05 | 4.9 | |
| MonkeyOCR-proSize=3B, Category=Specialized OCR2026.05 | 5.8 | |
| PA-BDMSize=3B2026.05 | 6.1 | |
| DODOSize=3B, Category=Diffusion VLMs2026.05 | 6.6 | |
| Mistral OCRCategory=Specialized OCR2026.05 | 7.2 | |
| Qwen 2.5 VLSize=72B, Category=Autoregressive VLMs2026.05 | 9.2 | |
| olmOCRSize=7B, Category=Specialized OCR2026.05 | 9.7 | |
| Nanonets-OCR-sSize=3B, Category=Specialized OCR2026.05 | 13.4 | |
| Qwen 2.5 VLSize=7B, Category=Autoregressive VLMs2026.05 | 13.5 | |
| DODO fastSize=3B, Category=Diffusion VLMs2026.05 | 15.9 | |
| Qwen 2.5 VLSize=3B, Category=Autoregressive VLMs2026.05 | 18.4 | |
| SmolDoclingSize=256M, Category=Specialized OCR2026.05 | 26.2 | |
| LLaDA-VSize=7B, Category=Diffusion VLMs2026.05 | 52.4 | |
| DimpleSize=7B, Category=Diffusion VLMs2026.05 | 85.6 | |
| LaViDa-LSize=8B, Category=Diffusion VLMs2026.05 | 99.4 |