PaddleOCR-VL vs MiniCPM-V 4.5
Pick PaddleOCR-VL when you need full-page document parsing on a modest GPU, with Markdown and JSON output, tables, formulas, charts and 109 languages. Pick MiniCPM-V 4.5 when you want one general 8B vision model for OCR plus document Q&A, including quantized builds for edge or on-device use. The main difference is size and focus: PaddleOCR-VL is a 0.9B document parser, MiniCPM-V 4.5 is an 8.7B general VLM.
PaddleOCR-VLpaddle/paddleocr-vl | MiniCPM-V 4.5openbmb/minicpm-v | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 6 credits / page | 2 credits / page |
| Per 1,000 pages | $6.00 | $2.00 |
| Typical speed | 1.3 s / page | 2.5 s / page |
| Best for | CJK documents | OCR on edge devices |
| Inputs | PDF, images | images |
| Languages | 109 languages | 30+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | ✓ Yes | No |
| Forms | No | No |
| Multilingual | ✓ Yes | ✓ 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 PaddleOCR-VL if
- It has under 1B parameters, so it runs on much smaller GPUs than 7B document VLMs.
- It recognizes text, tables, formulas and charts and outputs Markdown and JSON for full pages.
- It supports 109 languages, including Cyrillic, Arabic, Devanagari and Thai scripts, in one model.
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 is a general vision language model, so it fits local document Q&A on a single GPU as well as OCR.
- OpenBMB claims leading OCRBench results and state of the art PDF parsing among general MLLMs on OmniDocBench.
Key differences
| PaddleOCR-VL | MiniCPM-V 4.5 | |
|---|---|---|
| Model size | 0.9B parameters | 8.7B parameters |
| Licence | Apache-2.0 | Apache-2.0 for weights and code |
| How you run it | PaddleOCR, vLLM or transformers. The transformers path only does element-level recognition, not full page parsing. | llama.cpp, vLLM, Ollama or Transformers. Ollama and SGLang support were still pending official releases on the model card. |
| Hardware | NVIDIA GPU recommended. CPU inference is possible but slow. No minimum VRAM stated. | No int4 VRAM figure published. |
| Output | Markdown and JSON | Plain text, Markdown and JSON. No bounding box or layout output. |
| Benchmarks | OmniDocBench v1.6 overall 96.33 for version 1.6 (authors). olmOCR-Bench overall 80.0 (independent). | Scores are given only as images on the model card, so the numbers are hard to verify. Not published as text. |
| Known failures | Version 1.6 regresses Russian and Kazakh Cyrillic OCR on born-digital PDFs, decodes about 11x slower because use_cache is false, drops table content in some cases, and hangs on blurry images. | No specific failure reports listed. It cannot return coordinates. |
Sources
- PaddleOCR-VL Hugging Face model card
- PaddleOCR-VL-1.6 model card
- olmOCR-Bench table
- Issue #18088 Cyrillic regression
- MiniCPM-V 4.5 Hugging Face model card
- MiniCPM-V GitHub repo
What each one is
PaddleOCR-VL
Open-source document model from Baidu. Excellent on Chinese, Japanese and Korean.
PaddleOCR-VL 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": "paddle/paddleocr-vl","fallbacks": ["openbmb/minicpm-v"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
PaddleOCR-VL alternativesMiniCPM-V 4.5 alternativesBest OCR models in 2026