olmOCR vs Nanonets-OCR
Pick olmOCR for large-scale PDF to Markdown, academic papers with equations, and pages that mix print and handwriting. It scores 82.4 on olmOCR-Bench against 69.5 for Nanonets-OCR2-3B. Pick Nanonets-OCR for printed business documents with signatures, watermarks and checkboxes, or flowcharts you want as Mermaid, and when you have a GPU smaller than 12 GB.
olmOCRallenai/olmocr | Nanonets-OCRnanonets/nanonets-ocr | |
|---|---|---|
| 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 | 1.8 s / page |
| Best for | Bulk PDF to text for datasets | Markdown with semantic tags |
| Inputs | PDF, images | PDF, images |
| Languages | Mainly English | Multilingual |
| Max pages per request | 1,000 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | No | No |
| Confidence | No | No |
| Forms | No | ✓ Yes |
| Multilingual | No | ✓ 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 olmOCR if
- It scores 82.4 on olmOCR-Bench, while Nanonets-OCR2-3B scores 69.5 on the same benchmark.
- It handles equations, tables and handwriting, and it strips headers and footers in natural reading order.
- It is Apache-2.0 for the toolkit, with weights, training data and code released, and Ai2 puts self-hosted cost under $200 per million pages.
Choose Nanonets-OCR if
- It tags signatures, watermarks and image descriptions, and converts checkboxes to unicode symbols, which suits business forms.
- It extracts flowcharts and org charts as Mermaid code and can output HTML and LaTeX as well as Markdown.
- It is a 3B model that scores 89.43 on DocVQA, so it is the lighter option to self-host, and the weights are free.
Key differences
| olmOCR | Nanonets-OCR | |
|---|---|---|
| Model size | 7B | 3B (Hugging Face lists about 4B params) |
| Licence | Apache-2.0 (toolkit) | Not stated on the Nanonets-OCR2-3B model card; base model is Qwen2.5-VL-3B |
| How you run it | Self-host with vLLM on an NVIDIA GPU with at least 12 GB VRAM, or call third-party hosted endpoints | Self-host the weights from Hugging Face only |
| Price | Self-hosted under $200 per million pages (Ai2). DeepInfra hosted: $0.09 input / $0.19 output per million tokens | Free to self-host; you pay for compute |
| Output formats | Markdown, JSON | Markdown, HTML, LaTeX, Mermaid |
| olmOCR-Bench overall | 82.4 | 69.5 (OCR2-3B) |
| Handwriting | Handles handwriting, including mixed printed and handwritten pages | Not trained on handwriting and can hallucinate; one test saw generation loops with CER up to 2100% |
| Known failure | Out of memory on a 12 GB GPU, tends to get stuck on the last few pages, unreliable image bounding boxes | Hallucinates with prompts other than its fixed training prompt |
Sources
- olmOCR GitHub README
- Issue #424 OOM
- Issue #433 stuck
- Nanonets-OCR2-3B model card
- Nanonets-OCR-s announcement
- HF discussion fixed prompt
What each one is
olmOCR
Open-source vision model from Ai2 (often searched as olm ocr) that turns PDFs into clean, reading-order text. Latest release: olmOCR 2.
olmOCR 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": "allenai/olmocr","fallbacks": ["nanonets/nanonets-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
olmOCR alternativesNanonets-OCR alternativesBest OCR models in 2026