MDPBench

Multilingual document parsing benchmark covering digital and photographed documents across 17 languages and diverse scripts.

14rows
scoreprimary metric
2026-05-06sampled

Metadata

ID
mdpbench
Category
Multimodal
Release
2026-03-30

Metrics

MDPBench score

Latest Results

Rows are ranked by the Hugging Face leaderboard API rank. Model display names are preserved from source modelId values.

Rank Subject MDPBench score Model Match Provenance Sampled
1 rednote-hilab/dots.mocr 80.50 — Imported 2026-05-06
2 PaddlePaddle/PaddleOCR-VL-1.5 78.30 — Imported 2026-05-06
3 rednote-hilab/dots.ocr 76.50 — Imported 2026-05-06
4 allenai/olmOCR-7B-0225-preview 70.40 — Imported 2026-05-06
5 PaddlePaddle/PaddleOCR-VL 69.60 — Imported 2026-05-06
6 Qwen/Qwen3-VL-8B-Instruct 68.30 Qwen3 VL 8B Instruct
qwen-qwen3-vl-8b-instruct
 		Imported 2026-05-06
7 zai-org/GLM-OCR 67.30 — Imported 2026-05-06
8 Qwen/Qwen3.5-9B 65.70 Qwen3.5-9B
qwen-qwen3.5-9b
 		Imported 2026-05-06
9 nanonets/Nanonets-OCR2-3B 64.20 — Imported 2026-05-06
10 lightonai/LightOnOCR-2-1B 63.90 — Imported 2026-05-06
11 nanonets/Nanonets-OCR-s 63.70 — Imported 2026-05-06
12 deepseek-ai/DeepSeek-OCR 51.80 — Imported 2026-05-06
13 opendatalab/MinerU2.5-2509-1.2B 46.30 — Imported 2026-05-06
14 OpenGVLab/InternVL3_5-8B 42.70 — Imported 2026-05-06