MinerU vs Phi-4-multimodal
Pick MinerU for Chinese and English academic papers and mixed PDF and Office corpora, especially if you need CPU-only conversion or output beyond Markdown. Pick Phi-4-multimodal for English document Q&A on a modest GPU, where the MIT licence and 128K context help. The main split is task: MinerU is a document parser, Phi-4-multimodal is a general model that reads documents and audio.
MinerUopendatalab/mineru | Phi-4-multimodalmicrosoft/phi-4-multimodal | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.0 s / page | 2.5 s / page |
| Best for | Academic PDFs | Small model for OCR and questions |
| Inputs | PDF, images | images |
| Languages | 80+ languages | English for vision |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | ✓ Yes | No |
| 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 MinerU if
- You have Chinese or English academic papers, or a mix of PDF, DOCX, PPTX, XLSX, EPUB, OFD and HTML files to convert in one tool.
- You have no GPU: the basic ONNX tier runs on CPU with about 2 GB RAM.
- You need output as Markdown, HTML, LaTeX, DOCX, EPUB or JSON, not just text.
Choose Phi-4-multimodal if
- You need a permissive MIT licence for commercial use, with none of the AGPL-tagged weights or custom terms that MinerU carries.
- You want to ask questions about English documents: it scores 93.2 on DocVQA at 5.6B parameters, as reported by Microsoft.
- You also process audio, or want a 128K token context, in the same model.
Key differences
| MinerU | Phi-4-multimodal | |
|---|---|---|
| Licence | MinerU Open Source License (Apache-2.0 based with extra conditions); MinerU2.5 VLM weights tagged AGPL-3.0 | MIT |
| How you run it | Self-host only, as CLI, API or WebUI | Self-host, or use it through Azure AI Foundry |
| Hardware | Basic tier: CPU with 2 GB RAM. VLM tiers: 16 GB RAM and a GPU with 8 GB+ VRAM | Flash attention on A100, A6000 and H100; V100 and older need eager attention |
| Input types | PDF, images, DOCX, PPTX, XLSX, EPUB, OFD, HTML | Text, image and audio |
| Output formats | Markdown, HTML, LaTeX, DOCX, EPUB, JSON | Plain text, Markdown, JSON |
| Benchmarks | olmOCR-Bench overall 75.2 (MinerU 2.5.4); ParseBench mean 45.9 (MinerU2.5), both independent | DocVQA 93.2, OCRBench 84.4, ChartQA 81.4, InfoVQA 72.7, all reported by the authors |
| Known failure | VLM tiers drop merged header rows in tables such as bank statement account lines; repeated content can produce 300 KB figure blocks | Vision is English only and performance drops in other languages; the model card warns of repetitive output in long chats |
Sources
- MinerU GitHub README
- MinerU2.5 model card
- olmOCR-Bench table
- Issue #5554 table header rows
- Hugging Face model card
What each one is
MinerU
Open-source PDF to Markdown pipeline popular for academic and Chinese documents.
MinerU detailsPhi-4-multimodal
Microsoft's 5.6B text, vision and audio model (February 2025). Scores 84.4 on OCRBench per its model card.
Phi-4-multimodal 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": "opendatalab/mineru","fallbacks": ["microsoft/phi-4-multimodal"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
MinerU alternativesPhi-4-multimodal alternativesBest OCR models in 2026