OpenOCR vs MiniCPM-V 4.5
Pick OpenOCR for self-hosted Chinese and English document parsing, especially academic PDFs with formulas, when you want a 0.1B model and bounding boxes. Pick MiniCPM-V 4.5 for local or on-device OCR and document Q&A on a single GPU, with quantized builds and Markdown or JSON output. The main split is size and output: OpenOCR is tiny and gives coordinates, MiniCPM-V is an 8.7B general vision language model with no layout or box output.
OpenOCRfudan/openocr | MiniCPM-V 4.5openbmb/minicpm-v | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 1 credit / page | 2 credits / page |
| Per 1,000 pages | $1.00 | $2.00 |
| Typical speed | 0.5 s / page | 2.5 s / page |
| Best for | Lightweight OCR pipeline on CPU | OCR on edge devices |
| Inputs | PDF, images | images |
| Languages | Chinese and English | 30+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | No | ✓ Yes |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose OpenOCR if
- It reports 90.57 on OmniDocBench v1.5 (OpenDoc-0.1B) with only 0.1B parameters, though the authors measured it themselves.
- It returns bounding boxes and LaTeX, so you can get coordinates and formulas from academic PDFs.
- UniRec-0.1B reads plain text and single or multi-line formulas in one model, and the toolkit supports ONNX export.
Choose MiniCPM-V 4.5 if
- It runs on llama.cpp and vLLM with int4, GGUF and AWQ builds, which suits on-device or edge OCR.
- It outputs Markdown and JSON, and works as a general model for local document Q&A on a single GPU.
- It is not limited to Chinese and English by a stated language restriction, unlike OpenDoc and UniRec, and OpenBMB claims leading OCRBench results above GPT-4o-latest and Gemini 2.5.
Key differences
| OpenOCR | MiniCPM-V 4.5 | |
|---|---|---|
| Model size | 0.1B (OpenDoc-0.1B and UniRec-0.1B) | 8.7B, built on Qwen3-8B and SigLIP2-400M |
| Licence | Apache-2.0 | Apache-2.0 for weights and code |
| How you run it | Self-host as a Python library, with ONNX export | Self-host with llama.cpp, vLLM or Transformers; Ollama and SGLang support was still pending on the model card |
| Output formats | Plain text, LaTeX, bounding boxes | Plain text, Markdown, JSON; no bounding box or layout output |
| Languages | OpenDoc and UniRec support only Chinese and English | Not published |
| Benchmarks | 90.57 on OmniDocBench v1.5 (OpenDoc-0.1B), self-reported | Claims leading OCRBench and OmniDocBench PDF parsing results, but scores appear only as images on the model card |
| Known limits | Document parsing needs a separate layout model, PP-DocLayoutV2 from PaddlePaddle; no hardware or throughput figures | No int4 VRAM figure published; no hosted SLA |
Sources
What each one is
OpenOCR
Fudan's open toolkit for text detection and recognition, formulas, tables and document parsing.
OpenOCR detailsMiniCPM-V 4.5
8B vision model from OpenBMB (August 2025) built to run on devices, with strong OCR and document parsing.
MiniCPM-V 4.5 detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "fudan/openocr","fallbacks": ["openbmb/minicpm-v"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
OpenOCR alternativesMiniCPM-V 4.5 alternativesBest OCR models in 2026