OCRBench V2

A native OCR benchmark for reading text from images across multilingual scripts, low-quality scans, handwriting, structured layouts, charts, and screenshots.

5 models

Higher is better.

1Qwen3.8 MaxAlibaba74.2
2Qwen3.7 PlusAlibaba70.7
3Interfaze BetaInterfaze70.7
4Ternary Bonsai 2 27BPrism ML56.9
5LFM2.5-VL-3BLiquid AI47.5

Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric: Accuracy
Tasks: Image OCR tasks
Authors: OCRBench authors (2025)
Source: OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning ↗

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.