Marker vs Azure Document Intelligence
Pick Marker when you want to turn books, papers and other PDFs into Markdown, HTML or RAG chunks on your own hardware. Pick Azure Document Intelligence when you process invoices, receipts or ID documents and want prebuilt extraction models from a managed service. The main split is licence and cost: Marker is free to self-host but its weights restrict commercial use above $5M, while Azure is proprietary and billed per page.
Markerdatalab/marker | Azure Document Intelligenceazure/document-intelligence | |
|---|---|---|
| Type | OCR API | OCR API |
| Price | 14 credits / page | 60 credits / page |
| Per 1,000 pages | $14.00 | $60.00 |
| Typical speed | 2.5 s / page | 1.5 s / page |
| Best for | Books and research papers | Enterprise forms and IDs |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | 90+ languages | 300+ languages |
| Max pages per request | 1,000 | 2,000 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | ✓ Yes |
| Forms | No | ✓ Yes |
| 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 free self-hosted PDF to Markdown, and you only pay for your own compute.
- You need Markdown, HTML, JSON block trees, chunks or LaTeX output for RAG or for papers with equations.
- You have no GPU: fast mode runs on CPU with about 18 MB of ONNX weights at 23.6 pages per second.
Choose Azure Document Intelligence if
- You process invoices, receipts, IDs, W-2s or contracts, which have prebuilt models.
- Your company is above $5M and cannot use Marker's weights for free, or you already run on Azure.
- You need on-prem use with a managed vendor, through disconnected containers.
Key differences
| Marker | Azure Document Intelligence | |
|---|---|---|
| Licence | Code is Apache-2.0. Weights use a modified AI Pubs Open Rail-M licence, free for research, personal use and startups under $5M. | Proprietary. |
| How you run it | Local on GPU, CPU or Apple Silicon. Datalab also sells a hosted API. | Managed Azure service, or containers including disconnected ones. |
| Price | Free self-hosted within the weight licence. Hosted API has about 5,000 free pages a month. | Read is $1.50 per 1,000 pages ($0.60 above 1M). Layout and prebuilt are $10 per 1,000. Custom extraction is $30 per 1,000. |
| Free tier | About 5,000 free pages a month on the hosted API. | 500 free pages a month, and only the first two pages of each document are processed. |
| Output formats | Markdown, HTML, JSON, chunks, LaTeX, images. | JSON, Markdown, plain text, bounding boxes, searchable PDF. |
| Benchmarks | 83.7 on olmOCR-bench in vllm mode, 66.6 in CPU fast mode. These are Datalab's own numbers. | 82.7% table accuracy on RD-TableBench, from an independent test. The two scores are on different benchmarks. |
| Known cost trap | Commercial use above $5M needs a paid weight licence. | Stacking models adds up: Prebuilt plus Layout plus the high-res add-on is $26 per 1,000 pages. Batch costs the same as synchronous calls. |
Sources
- Marker GitHub README
- Datalab pricing
- Azure pricing page
- Azure DI pricing breakdown
- Reducto RD-TableBench
What each one is
Marker
Hosted version of the open-source Marker and Surya models. Strong on books and papers.
Marker detailsAzure Document Intelligence
Azure layout model with tables, selection marks and prebuilt invoice and ID readers.
Azure Document Intelligence 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": ["azure/document-intelligence"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Marker alternativesAzure Document Intelligence alternativesBest OCR models in 2026