Marker vs Surya
Pick Marker if you want PDFs converted to Markdown, HTML, JSON or RAG chunks in one step, including books and papers with equations. Pick Surya if you need layout labels, reading order and bounding boxes for 90+ languages from a lightweight local client. Marker is the converter built on the Surya models, so the choice is mostly about output: finished documents versus block-level OCR data.
Markerdatalab/marker | Suryadatalab/surya | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 14 credits / page | 2 credits / page |
| Per 1,000 pages | $14.00 | $2.00 |
| Typical speed | 2.5 s / page | 1.5 s / page |
| Best for | Books and research papers | Layout and line-level OCR |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | 90+ languages | 90+ languages |
| Max pages per request | 1,000 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| 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 Marker if
- You want a finished PDF to Markdown conversion, with HTML, JSON block trees, LaTeX and RAG chunks as outputs.
- You need a CPU option: fast mode runs on CPU at 23.6 pages per second with about 18 MB of ONNX weights, though it scores 66.6 on olmOCR-bench.
- You can use the hosted Datalab API, which includes about 5,000 free pages per month, instead of running it yourself.
Choose Surya if
- You need line and block level bounding boxes, layout labels and reading order from one model.
- You want local OCR on a laptop: it runs on Apple Silicon or CPU through llama.cpp, with peak memory of about 4.3 to 5.9 GB on an M4 Max.
- You work in many languages and need OCR for 90+ of them, and can accept that only 21 of 91 languages reach a 90% pass rate.
Key differences
| Marker | Surya | |
|---|---|---|
| What it is | PDF to Markdown converter built on the Surya OCR models | OCR, layout analysis, reading order and table recognition toolkit |
| How you run it | Locally on GPU, CPU or Apple Silicon, or through Datalab's hosted API | Self-hosted only, as a lightweight client against a vLLM or llama.cpp server you run separately |
| Outputs | Markdown, HTML, JSON, chunks, LaTeX, images | JSON, HTML, LaTeX, bounding boxes, plain text |
| olmOCR-bench (author-reported) | 83.7 in vllm mode; 66.6 in CPU fast mode | 83.7% for Surya 2.5 |
| Licence | Code Apache-2.0; weights under modified AI Pubs Open Rail-M, free for research, personal use and startups under $5M | Code Apache-2.0; weights under modified AI Pubs Open Rail-M, free for research, personal use and companies under $5M funding or revenue |
| Price | Free self-hosted (you pay for compute); hosted API has about 5,000 free pages per month | Free self-hosted (you pay for compute); commercial licence needed above $5M revenue or funding |
| Known failure | Weights block free commercial use for companies above $5M; all benchmark numbers are Datalab's own | Surya 2.5 table output lacks row/column geometry; multilingual tables score 58.1; not for handwriting or photos |
Sources
What each one is
Marker
Hosted version of the open-source Marker and Surya models. Strong on books and papers.
Marker detailsSurya
Open-source OCR toolkit with line detection, layout analysis, reading order and tables.
Surya 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/marker","fallbacks": ["datalab/surya"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Marker alternativesSurya alternativesBest OCR models in 2026