Nanonets-OCR vs Pixtral
Pick Nanonets-OCR if you self-host and want printed business PDFs turned into Markdown, with signatures, watermarks, checkboxes and flowcharts (as Mermaid) tagged. Pick Pixtral only if you already run on the Mistral API or need question answering over charts and documents. Pixtral is marked deprecated and no longer maintained, so it is a poor base for a new OCR project.
Nanonets-OCRnanonets/nanonets-ocr | Pixtralmistral/pixtral | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 1.8 s / page | 3.0 s / page |
| Best for | Markdown with semantic tags | Questions about document images |
| Inputs | PDF, images | images |
| Languages | Multilingual | Multilingual |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | No |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | No | No |
| Confidence | No | No |
| Forms | ✓ Yes | 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 Nanonets-OCR if
- You need structured output: it tags signatures, watermarks and image descriptions, turns checkboxes into unicode symbols, and writes flowcharts and org charts as Mermaid.
- You want Markdown, HTML or LaTeX output from printed PDFs, including inline and display equations told apart.
- You want a small 3B model you host yourself at no licence fee, paying only for compute.
Choose Pixtral if
- You already have an integration on the Mistral API, where launch pricing was $0.15 per million input and output tokens.
- You need question answering over charts and documents: DocVQA is 90.7 ANLS and ChartQA is 81.8, per Mistral.
- You want Apache-2.0 open weights and a 128K token context that fits many images in one request.
Key differences
| Nanonets-OCR | Pixtral | |
|---|---|---|
| Licence | Not stated on the model card (no license tag) | Apache-2.0 |
| How you run it | Self-host the weights from Hugging Face | Mistral API (pixtral-12b-2409) or self-host open weights |
| Price | Free to self-host, you pay for compute | $0.15 per million input and output tokens at API launch; free to self-host |
| Output formats | Markdown, HTML, LaTeX, Mermaid | Plain text, Markdown, JSON |
| DocVQA | 89.43 (OCR2-3B, model card) | 90.7 ANLS (Mistral) |
| Handwriting | Not trained on it; can hallucinate, with generation loops and CER up to 2100% in one test | Reported accuracy drop on handwriting, small fonts and non-Latin scripts |
| Known issue | Built on a fixed prompt; other prompts are expected to hallucinate. olmOCR-bench overall is 69.5 | Deprecated and no longer maintained; one HN user said it does not always get OCR near correct |
Sources
- Nanonets-OCR2-3B model card
- Nanonets-OCR-s announcement
- HF discussion fixed prompt
- Mistral announcement
- Mistral Sept 2024 pricing
- HN launch thread
What each one is
Nanonets-OCR
Open image-to-Markdown model that tags tables, equations, signatures, watermarks and checkboxes.
Nanonets-OCR detailsPixtral
Mistral's open-weight vision models, Pixtral 12B and Pixtral Large. Mistral's API no longer serves them, so they run self-hosted.
Pixtral 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": "nanonets/nanonets-ocr","fallbacks": ["mistral/pixtral"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Nanonets-OCR alternativesPixtral alternativesBest OCR models in 2026