OCR models / Huazhong University of Science and Technology

MonkeyOCR-pro 3B

Name: MonkeyOCR-pro 3B
Author: Huazhong University of Science and Technology

Huazhong University of Science and Technology · released 2025-07-10 · data checked 2026-11-06

MonkeyOCR-pro 3B is a 3.8B-parameter OCR and document-parsing model from Huazhong University of Science and Technology (CN), released 2025-07-10 under the Non-commercial license. It scores 88.57 on OmniDocBench v1.6 (#30 of 33) and 75.8 on olmOCR-Bench (#17 of 24). It needs about 8.5 GB of memory at 16-bit.

Layout-then-recognize pipeline built on Qwen2.5-VL 3B. Succeeded by MonkeyOCR v2.

OmniDocBench v1.6

88.57

#30 of 33 models

olmOCR-Bench

75.8

#17 of 24 models

Parameters

3.8B

OCR specialist

Memory to run

~8.5 GB

16-bit weights plus about 1 GB

Benchmarks

OmniDocBench v1.6 Leaderboard88.57

Best: OXR 1.0 97.03

olmOCR-Bench Lab-reported75.8

Best: OXR 1.0 87.8

The tick marks the best score on each benchmark. See how to read the ranking.

Specifications

Lab: Huazhong University of Science and Technology (CN)
Type: OCR / document parsing specialist
Total parameters: 3.8B
License: Non-commercial
The card shows an Apache 2.0 badge but says the model is for academic research and non-commercial use only.
Memory to run: ~8.5 GB at 16-bit
Released: 2025-07-10

Weights and usage on Hugging Face ↗

Closest alternatives

OCRVerse (4.8B, 88.60 on OmniDocBench, Not stated)
Dolphin v2 (3.8B, 89.50 on OmniDocBench, Not stated)
HunyuanOCR 1.0 (1B, 89.95 on OmniDocBench, Tencent Hunyuan Community License)
MiniCPM-V 4.6 (1.3B, 86.93 on OmniDocBench, Apache-2.0)

Sources

Something wrong here? Report it.