Marker vs Nanonets-OCR
Pick Marker for self-hosted PDF to Markdown on books and papers with equations, or when you want RAG chunks, and check its weight license first if your company is above $5M. Pick Nanonets-OCR for printed business documents with signatures, watermarks and checkboxes, or flowcharts you want as Mermaid. The main gap is accuracy and licensing: Marker scores 83.7 on olmOCR-bench in vllm mode against 69.5 for Nanonets-OCR2-3B, but Nanonets-OCR has no stated license.
Markerdatalab/marker | Nanonets-OCRnanonets/nanonets-ocr | |
|---|---|---|
| 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 | 1.8 s / page |
| Best for | Books and research papers | Markdown with semantic tags |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | 90+ languages | Multilingual |
| Max pages per request | 1,000 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | No |
| 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
- Its vllm mode scores 83.7 on olmOCR-bench, versus 69.5 for Nanonets-OCR2-3B (both are the authors' own numbers).
- It runs on CPU, GPU or Apple Silicon, and fast mode needs only about 18 MB of ONNX weights and does 23.6 pages per second on CPU.
- It outputs Markdown, HTML, JSON block trees, chunks, LaTeX and images, and Datalab offers a hosted API with about 5,000 free pages per month.
Choose Nanonets-OCR if
- It tags signatures, watermarks and image descriptions, and turns checkboxes into unicode symbols, which suits forms and business documents.
- It extracts flowcharts and org charts as Mermaid code, and separates inline from display LaTeX equations.
- It is a single 3B model you self-host from Hugging Face with no paid tier or weight-license revenue cap stated, and it scores 89.43 on DocVQA.
Key differences
| Marker | Nanonets-OCR | |
|---|---|---|
| License | Code Apache-2.0; weights under a modified AI Pubs Open Rail-M license, free for research, personal use and startups under $5M. | Not stated on the Nanonets-OCR2-3B model card. |
| How you run it | Self-host on GPU, CPU or MPS, or use Datalab's hosted API. vllm mode needs an NVIDIA GPU. | Self-host only. Fine-tuned from Qwen2.5-VL-3B, weights on Hugging Face. |
| Price | Free self-hosted within the weight license. 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. | Markdown, HTML, LaTeX, Mermaid. |
| olmOCR-bench overall | 83.7 in vllm mode. 66.6 in CPU fast mode. | 69.5 for OCR2-3B. |
| Handwriting | Not published | Not trained on handwriting and can hallucinate. One test saw generation loops with CER up to 2100% and 69.1s per image on an L4. |
| Prompts | Not published | Trained on a fixed prompt. The team says hallucination with other prompts is expected. |
Sources
- Marker GitHub README
- Datalab pricing
- Nanonets-OCR2-3B model card
- Nanonets-OCR-s announcement
- HF discussion fixed prompt
- Japanese handwriting OCR comparison
What each one is
Marker
Hosted version of the open-source Marker and Surya models. Strong on books and papers.
Marker detailsNanonets-OCR
Open image-to-Markdown model that tags tables, equations, signatures, watermarks and checkboxes.
Nanonets-OCR 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": ["nanonets/nanonets-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Marker alternativesNanonets-OCR alternativesBest OCR models in 2026