InduOCRBench: leaderboard

InduOCRBench (2026) collects 570 real industrial documents, 3,402 pages in 11 challenge types such as ultra-wide Gantt charts, receipts, historical books, watermarks and cross-page tables, and scores OCR and document-parsing systems on their structured Markdown output. The overall score is the mean of text edit-distance similarity, formula CDM and table TEDS, 0 to 100; higher is better.

Metric: Average OCR score (0-100): mean of the text edit-distance similarity, formula CDM and table TEDS scores on 570 industrial documents (3,402 pages) in 11 challenge types (complex background, high pixel count, ultra-long, multi-column, ultra-wide, historical books, handwriting, multi-font, visual style, watermark, cross-page tables) plus normal documents, each OCR method converting pages to structured Markdown (general VLMs with a standardized prompt, pipeline tools at default configuration, closed solutions through their official APIs); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1PaddleOCR-VL78.2
2Gemini 2.5 Pro74.5#276
3MinerU2.5-2509-1.2B72.5
4Qwen 3 VL 235B A22B Instruct70.9#532
5HunyuanOCR68.1
6Mineru2-pipeline66.5
7Doc2X61.6
8DeepSeek-OCR61.5
9PP-StructureV360.3
10GPT-4o52#751

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=induocrbench · How It Works · Data refreshed daily, snapshot 2026-11-21.