MinerU vs Qwen-VL
Pick MinerU for batch conversion of Chinese and English papers and mixed PDF and Office files, especially if you have no GPU. Pick Qwen-VL for photos, screenshots and jobs that mix OCR with reasoning, and add loop checks on long runs. MinerU scores higher on olmOCR-Bench (75.2 vs 64.6 for Qwen 3 VL 8B), and it is a document parser, while Qwen-VL is a general model.
MinerUopendatalab/mineru | Qwen-VLqwen/qwen-vl | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.0 s / page | 2.0 s / page |
| Best for | Academic PDFs | Flexible extraction with prompts |
| Inputs | PDF, images | PDF, images |
| Languages | 80+ languages | 30+ languages |
| Max pages per request | 500 | 200 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | 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 MinerU if
- It scores 75.2 on olmOCR-Bench (MinerU 2.5.4), above the 64.6 that Qwen 3 VL 8B gets.
- One tool takes PDF, images, DOCX, PPTX, XLSX, EPUB, OFD and HTML, and renders to Markdown, HTML, LaTeX, DOCX, EPUB or JSON.
- The basic tier runs on CPU with about 2 GB RAM, and the VLM tier reaches 2.12 pages per second on one A100 with vLLM.
Choose Qwen-VL if
- It does OCR in 32 languages and handles low light, blur and tilted text, so it suits photos and screenshots with text.
- It is Apache-2.0 in every size from 2B to 235B-A22B MoE, with no extra licence terms, and it can also be used as a hosted API on Alibaba Cloud.
- It returns bounding boxes and a Qwen HTML document format, and it can answer questions about the page, so one model covers OCR plus reasoning or VQA.
Key differences
| MinerU | Qwen-VL | |
|---|---|---|
| Licence | MinerU Open Source License (Apache-2.0 based with extra conditions); MinerU2.5 VLM weights tagged AGPL-3.0 | Apache-2.0 (Qwen3-VL repository) |
| How you run it | Self-host only: CLI, API or WebUI | Self-host the open weights, or use the hosted API on Alibaba Cloud |
| Hardware | Basic tier: CPU, 2 GB RAM. VLM tiers: 16 GB RAM and a GPU with 8 GB+ VRAM | FP8 checkpoints need an H100-class GPU and CUDA 12+; the 235B example uses 8-way tensor parallel on vLLM |
| Model size | MinerU2.5 VLM: 1.2B | 2B, 4B, 8B, 32B dense; 30B-A3B and 235B-A22B MoE |
| Output formats | Markdown, HTML, LaTeX, DOCX, EPUB, JSON | Plain text, HTML, bounding boxes, JSON, Markdown |
| olmOCR-Bench | 75.2 overall (MinerU 2.5.4) | 64.6 (Qwen 3 VL 8B) |
| Known failure | Repeated content such as 300 KB figure blocks (issue #5728); merged header rows dropped in tables such as bank statements (issue #5554); full-width Chinese punctuation turned half-width with no switch (issue #5765) | Loops repeating the same text, worse with greedy decoding (issue #1611) and on low-quality scans; Qwen2.5-VL-32B GGUF often gets stuck on scanned documents (llama.cpp #14663) |
Sources
- MinerU GitHub README
- olmOCR-Bench table
- Issue #5554 table header rows
- Qwen3-VL GitHub
- Issue #1611 infinite repetition
- Chandra README benchmark table
What each one is
MinerU
Open-source PDF to Markdown pipeline popular for academic and Chinese documents.
MinerU detailsQwen-VL
Alibaba's open-weight vision language models (Qwen2.5-VL, Qwen3-VL). Strong general OCR and grounding.
Qwen-VL 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": ["qwen/qwen-vl"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
MinerU alternativesQwen-VL alternativesBest OCR models in 2026