InduOCRBench - Text: leaderboard

The text recognition score of InduOCRBench (2026), a document-parsing benchmark of 570 real industrial documents in 11 challenge types: one minus the normalized edit distance of the recognized text, times 100. Higher is better.

Metric: Text edit-distance similarity (0-100): one minus the normalized edit distance between the recognized and the reference text, times 100, 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-VL88.1
2HunyuanOCR86.1
3Qwen 3 VL 235B A22B Instruct83.3#532
4Gemini 2.5 Pro83.1#276
5MinerU2.5-2509-1.2B81.8
6Mineru2-pipeline80.1
7PP-StructureV378.2
8Doc2X76.5
9DeepSeek-OCR75.5
10GPT-4o60.8#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-text · How It Works · Data refreshed daily, snapshot 2026-11-21.