Mistral OCR vs SmolDocling
Pick Mistral OCR if you want a hosted API for high-volume PDF to Markdown at $2 per 1,000 pages ($1 via Batch API). Pick SmolDocling only for English experiments on your own hardware, since it is a 256M preview model that its authors say is superseded by Granite-Docling. The main split is a paid, proprietary API against a free, self-hosted preview.
Mistral OCRmistral/mistral-ocr | SmolDoclinghuggingface/smoldocling | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 25 credits / page | 1 credit / page |
| Per 1,000 pages | $25.00 | $1.00 |
| Typical speed | 0.9 s / page | 1.0 s / page |
| Best for | High-volume PDF to Markdown | Very cheap page conversion |
| Inputs | PDF, images | PDF, images |
| Languages | 100+ languages | Mainly English |
| Max pages per request | 1,000 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| 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 Mistral OCR if
- It is a hosted API with no GPU to run, priced at $2 per 1,000 pages and $1 per 1,000 through the Batch API.
- OCR 3 rebuilds tables as HTML with rowspan and colspan, and the API returns Markdown, HTML, JSON and images.
- Mistral reports a 74% win rate for OCR 3 over OCR 2 on forms, scans, tables and handwriting.
Choose SmolDocling if
- It is free to use under Apache-2.0 per the model summary, and you pay only for your own compute.
- At 256M parameters it runs on modest hardware, and the authors report about 0.35 seconds per page on an A100 with vLLM.
- It outputs DocTags plus Markdown, HTML, JSON, LaTeX and bounding boxes, and one model covers text, tables, code, formulas and charts.
Key differences
| Mistral OCR | SmolDocling | |
|---|---|---|
| How you run it | Hosted API (pinned version mistral-ocr-2512, OCR 3) | Self-hosted, 256M parameters |
| Licence | Proprietary | Apache-2.0 per model summary (metadata tag says CDLA-Permissive-2.0) |
| Price | $2 per 1,000 pages; $3 with structured annotations; $1 via Batch API | Free, you pay for compute |
| Output formats | Markdown, HTML, JSON, images | DocTags, Markdown, HTML, JSON, LaTeX, bounding boxes |
| Languages | Not published | English only |
| Benchmark | OmniDocBench v1.6 overall 85.66; v1.5 score for the first version (OCR 2503) was 78.83 | Table TEDS 0.82 on FinTabNet, full-page OCR F1 0.80 (author-reported) |
| Known failure | Hallucinated text on unreadable handwriting; colored-background pages returned as images in the 2025 release; changed checkbox states and misread characters on a dense financial table | Endless token repetition and missing closing tags; output loops on some images, making generation very slow (Docling issue #1515) |
Sources
- Mistral OCR 3 announcement
- VentureBeat on OCR 3 pricing
- OmniDocBench v1.6 leaderboard
- SmolDocling model card
- Docling issue #1515
- Granite-Docling comparison table
What each one is
Mistral OCR
Fast document OCR that returns clean Markdown with tables and embedded images.
Mistral OCR detailsSmolDocling
Tiny 256M vision model from Hugging Face and IBM that outputs DocTags. Predecessor of Granite Docling.
SmolDocling 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": "mistral/mistral-ocr","fallbacks": ["huggingface/smoldocling"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Mistral OCR alternativesSmolDocling alternativesBest OCR models in 2026