Surya vs MonkeyOCR
Pick Surya for local layout detection, reading order and bounding boxes across 90+ languages, especially on a laptop or CPU. Pick MonkeyOCR for English and Simplified Chinese PDFs such as academic papers with formulas and tables, but only for research unless you get a written commercial license. Surya is free for companies under $5M funding or revenue, while MonkeyOCR v1 weights are non-commercial.
Suryadatalab/surya | MonkeyOCRmonkeyocr/monkeyocr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 1.5 s / page | 1.5 s / page |
| Best for | Layout and line-level OCR | Chinese and English PDFs |
| Inputs | PDF, images | PDF, images |
| Languages | 90+ languages | Chinese, English |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | No | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| 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 Surya if
- It runs on CPU and Apple Silicon through llama.cpp, with peak memory of about 4.3-5.9 GB on an M4 Max.
- One 689M model returns layout labels, reading order, OCR, HTML tables and KaTeX math with per-block bounding boxes.
- It covers 90+ languages and is free for personal use and for companies under $5M funding or revenue.
Choose MonkeyOCR if
- Pro-3B reports the best overall OmniDocBench edit distance among the models compared, in English and Chinese (0.138 / 0.206).
- It reports table TEDS of 81.5 English and 87.5 Chinese on OmniDocBench, which suits papers with tables and formulas.
- Pro-1.2B is about 36% faster than pro-3B for about 1.6% lower accuracy, and quantized weights run on an 8GB RTX 4060.
Key differences
| Surya | MonkeyOCR | |
|---|---|---|
| Licence | Code Apache-2.0; weights free for research, personal use and companies under $5M funding/revenue | Code Apache-2.0; v1 weights research and non-commercial only, commercial use needs a written license |
| How you run it | Light client against a vLLM or llama.cpp server you run yourself | Self-hosted via LMDeploy or Docker; 20/30/40-series and V100 Docker needs a patched image |
| Model size | 689M parameters | 3B and 1.2B |
| Hardware | NVIDIA GPU via vLLM, or CPU/Apple Silicon via llama.cpp | Tested on RTX 3090/4090, A6000, H800, A100; RTX 4060 8GB with quantized models |
| Speed | 4.52 pages/s at 192 DPI on one H100 with vLLM; 0.198 pages/s on an M4 Max at 96 DPI with llama.cpp Q8_0 | About 0.49 pages/s for pro-3B on a 3090 |
| Output formats | JSON, HTML, LaTeX, bounding boxes, plain text | Markdown, JSON, LaTeX, HTML, bounding boxes |
| Known failure | Surya 2.5 table output lacks row/column geometry; multilingual tables score 58.1; not for photos or natural scenes | Photographed text, handwriting, Traditional Chinese and multilingual text are not fully supported |
Sources
What each one is
Surya
Open-source OCR toolkit with line detection, layout analysis, reading order and tables.
Surya detailsMonkeyOCR
Open-source document parser that splits work into structure, recognition and relation steps.
MonkeyOCR 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": "datalab/surya","fallbacks": ["monkeyocr/monkeyocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Surya alternativesMonkeyOCR alternativesBest OCR models in 2026