Mistral OCR vs Chandra
Pick Mistral OCR for high-volume PDF to Markdown and batch scans of business documents where a hosted API at $1 to $2 per 1,000 pages is enough. Pick Chandra for multilingual scans, layout with bounding boxes, or when you need to self-host on your own GPU. The main split is hosted-only proprietary versus open weights you can run yourself (with a licence limit above $2M revenue or funding).
Mistral OCRmistral/mistral-ocr | Chandradatalab/chandra | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 25 credits / page | 2 credits / page |
| Per 1,000 pages | $25.00 | $2.00 |
| Typical speed | 0.9 s / page | 2.0 s / page |
| Best for | High-volume PDF to Markdown | Complex layouts and handwriting |
| Inputs | PDF, images | PDF, images |
| Languages | 100+ languages | 40+ languages |
| Max pages per request | 1,000 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| 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 Mistral OCR if
- Its list price is $2 per 1,000 pages, and the Batch API halves OCR to $1 per 1,000 pages.
- OCR 3 rebuilds tables as HTML with rowspan and colspan.
- It is a single hosted API with no GPU to run, and Mistral reports a 74% win rate for OCR 3 over OCR 2 on forms, scans, tables and handwriting.
Choose Chandra if
- You can self-host the open weights with vLLM on one NVIDIA GPU, or use the hosted Datalab API.
- It returns layout chunks with bounding boxes and labels per block, and scores 77.8% on a 43-language average.
- Hosted pricing starts at $1 per 1,000 pages (fast tier), and a free tier gives $20 per month of usage with a work email.
Key differences
| Mistral OCR | Chandra | |
|---|---|---|
| How you run it | Hosted Mistral API only | Self-host (vLLM or Hugging Face) or hosted Datalab API |
| Licence | Proprietary | Code Apache-2.0; weights modified OpenRAIL-M, free for research, personal use and companies under $2M funding/revenue |
| Price per 1,000 pages | $2; $3 with structured annotations; $1 via Batch API | Hosted: $1 (fast), $3 (balanced) or $7.50 (accurate); self-hosted: free within licence terms plus compute |
| Output formats | Markdown, HTML, JSON, images | Markdown, HTML, JSON, bounding boxes |
| Benchmark | OmniDocBench v1.6 overall 85.66 (independent); first version scored 78.83 on v1.5 | olmOCR-bench overall 85.8 for Chandra 2 (reported by the authors) |
| Languages | Not published | 77.8% average across 43 languages; 72.7% across 90 languages; 43.2% Urdu and 51.1% Sanskrit |
| Known failure | Hallucinated text on unreadable handwriting; colored-background pages returned as images with no OCR text in the 2025 release; changed checkbox states on a dense financial table | Token-loop hallucinations on 5-6 of about 60 Russian/Kazakh scans; Datalab says raw output hallucinates on about 6 in 10 of its hardest pages before the hosted pipeline's checks; Old Scans only 51.1 on olmOCR-bench |
Sources
- Mistral OCR 3 announcement
- OmniDocBench v1.6 leaderboard
- Reducto test of Mistral OCR
- Chandra GitHub README
- Datalab pricing
- GitHub issue #51 token loops
What each one is
Mistral OCR
Fast document OCR that returns clean Markdown with tables and embedded images.
Mistral OCR detailsChandra
Datalab's OCR model that keeps layout and handles forms, tables and handwriting.
Chandra 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": ["datalab/chandra"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Mistral OCR alternativesChandra alternativesBest OCR models in 2026