Qianfan OCR vs MinerU
Pick Qianfan OCR for self-hosted multilingual parsing under a plain Apache-2.0 licence, with 93.12 on OmniDocBench v1.5 and 79.8 on olmOCR-Bench. Pick MinerU when you need to ingest Office files and PDFs in one tool, run on CPU, or get higher throughput (2.12 pages per second on one A100 with vLLM). Watch MinerU's custom licence and its table header problems on bank statements.
Qianfan OCRbaidu/qianfan-ocr | MinerUopendatalab/mineru | |
|---|---|---|
| Type | OCR API | 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 | Chinese documents | Academic PDFs |
| Inputs | PDF, images | PDF, images |
| Languages | Chinese, English and more | 80+ languages |
| Max pages per request | 200 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | No | ✓ 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 Qianfan OCR if
- Its Apache-2.0 licence allows commercial use with no extra conditions.
- It covers 192 languages, handwriting and scene text, and suits key information extraction.
- It scores 79.8 on olmOCR-Bench, against 75.2 for MinerU 2.5.4 in the independent olmOCR table.
Choose MinerU if
- It accepts PDF, images, DOCX, PPTX, XLSX, EPUB, OFD and HTML in one tool.
- The basic ONNX tier runs on CPU with about 2 GB RAM, so you can start without a GPU.
- The VLM tier reaches 2.12 pages per second on one A100 with vLLM, about twice Qianfan OCR's 1.024.
Key differences
| Qianfan OCR | MinerU | |
|---|---|---|
| Licence | Apache-2.0 | MinerU Open Source License (Apache-2.0 based with extra conditions); VLM weights tagged AGPL-3.0 |
| How you run it | Self-host the weights or use the Baidu Qianfan platform | Self-host only, as a CLI, API or WebUI |
| Model size and hardware | 4B parameters; benchmarked on one A100 with vLLM; CPU-only deployment is not a fit | MinerU2.5 VLM is 1.2B; VLM tiers need 16 GB RAM and a GPU with 8 GB+ VRAM; basic tier runs on CPU with 2 GB RAM |
| Output formats | Markdown, JSON, HTML, LaTeX, bounding boxes | Markdown, HTML, LaTeX, DOCX, EPUB, JSON |
| Speed on an A100 | 1.024 pages per second, even with W8A8 | 2.12 pages per second with vLLM (VLM tier) |
| Benchmark | olmOCR-Bench 79.8 (authors); OmniDocBench v1.5 overall 93.12 (authors) | olmOCR-Bench 75.2 for MinerU 2.5.4 (independent); ParseBench mean 45.9 (independent) |
| Known failure | Thinking mode can emit up to 16,384 tokens, adding latency | VLM tiers drop merged header rows in tables such as bank statement account lines; the VLM pipeline can generate repeated content, including 300 KB figure blocks |
Sources
- Qianfan-OCR model card
- MinerU GitHub README
- MinerU2.5 model card
- olmOCR-Bench table
- Issue #5554 table header rows
What each one is
Qianfan OCR
Baidu's OCR-tuned vision model. Strong on Chinese documents and tables.
Qianfan OCR detailsMinerU
Open-source PDF to Markdown pipeline popular for academic and Chinese documents.
MinerU 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": "baidu/qianfan-ocr","fallbacks": ["opendatalab/mineru"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Qianfan OCR alternativesMinerU alternativesBest OCR models in 2026