Marker vs RolmOCR
Pick Marker if you want PDF to Markdown with HTML, JSON or RAG chunks, especially for books and papers with equations, and your company is under the $5M weight-license limit or will pay for a license. Pick RolmOCR if you need Apache-2.0 weights with no commercial limits and only want plain text from scanned pages, including rotated ones. The main split is output and licence: Marker gives structured output under a restricted weight licence, RolmOCR gives plain text under a free one.
Markerdatalab/marker | RolmOCRreducto/rolmocr | |
|---|---|---|
| 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 | 3.0 s / page |
| Best for | Books and research papers | Fast open PDF to text |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | 90+ languages | Not stated by the authors |
| Max pages per request | 1,000 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | No |
| 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 Marker if
- You need Markdown, HTML, JSON block trees, LaTeX or RAG chunks, not just plain text.
- You have no GPU: fast mode runs on CPU at 23.6 pages per second with about 18 MB of ONNX weights, though its olmOCR-bench score drops to 66.6.
- You want a hosted option: Datalab's API includes about 5,000 free pages per month.
Choose RolmOCR if
- You need fully free commercial use, since the weights are Apache-2.0 with no revenue cap.
- You process scanned pages that are rotated or skewed, because about 15% of its training data was rotated.
- You run one GPU and want a simple self-hosted plain text pipeline, for example with vLLM.
Key differences
| Marker | RolmOCR | |
|---|---|---|
| Licence | Code Apache-2.0; weights under a modified AI Pubs Open Rail-M licence, free for research, personal use and startups under $5M | Apache-2.0, usable commercially |
| How you run it | Self-host on GPU, CPU or Apple Silicon, or use Datalab's hosted API | Self-host only, for example with vLLM |
| Price | Free self-hosted within the weight licence; hosted API has about 5,000 free pages per month, then paid plans | Free self-hosted; you pay for compute |
| Output formats | Markdown, HTML, JSON, chunks, LaTeX, images | Plain text only; no layout bounding boxes |
| Benchmark | 83.7 on olmOCR-bench in vllm mode (Datalab's own number) | 0.663 overall on the Needl.ai benchmark (independent), versus 0.770 for olmocr-2-7b |
| Hardware | GPU, CPU or MPS; vllm mode needs an NVIDIA GPU | 8B model based on Qwen2.5-VL-7B, so it needs a GPU |
| Known failure | CPU fast mode drops to 66.6 on olmOCR-bench | Can hallucinate or drop content; Reducto shows it omitting a table subtitle; quantized versions not evaluated |
Sources
What each one is
Marker
Hosted version of the open-source Marker and Surya models. Strong on books and papers.
Marker detailsRolmOCR
Reducto's open PDF-to-text model, built on Qwen2.5-VL-7B with the olmOCR recipe. Needs no PDF metadata and handles rotated pages.
RolmOCR 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": ["reducto/rolmocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Marker alternativesRolmOCR alternativesBest OCR models in 2026