Chandra vs OpenOCR
Pick Chandra for multilingual scans, or when you need Markdown, HTML or JSON with layout boxes, or a hosted API. Pick OpenOCR for Chinese and English documents, especially academic PDFs with formulas, when you want a 0.1B model under Apache-2.0 with no revenue limit. The main split is language coverage and licence: Chandra covers many languages but has a revenue cap on its weights, while OpenOCR handles only Chinese and English under plain Apache-2.0.
Chandradatalab/chandra | OpenOCRfudan/openocr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 1 credit / page |
| Per 1,000 pages | $2.00 | $1.00 |
| Typical speed | 2.0 s / page | 0.5 s / page |
| Best for | Complex layouts and handwriting | Lightweight OCR pipeline on CPU |
| Inputs | PDF, images | PDF, images |
| Languages | 40+ languages | Chinese and English |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | No |
| Forms | ✓ Yes | 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 Chandra if
- You need many languages: Chandra 2 averages 77.8% across 43 languages and 72.7% across 90 languages.
- You want structured output: Markdown, HTML or JSON, plus layout chunks with bounding boxes and labels per block.
- You want the choice of a hosted API: Datalab convert costs $1, $3 or $7.50 per 1,000 pages, with $20/month free.
Choose OpenOCR if
- Your documents are Chinese or English, such as academic PDFs with formulas, and you want LaTeX output.
- You want a small model: OpenDoc-0.1B and UniRec-0.1B are 0.1B parameters, and OpenDoc-0.1B reports 90.57 on OmniDocBench v1.5.
- You need free commercial use at any company size: the licence is plain Apache-2.0, with no revenue or funding cap.
Key differences
| Chandra | OpenOCR | |
|---|---|---|
| Licence | Code Apache-2.0; weights modified OpenRAIL-M, free under $2M funding/revenue, commercial licence needed above that | Apache-2.0 |
| How you run it | Self-host with vLLM or Hugging Face, or use the hosted Datalab API | Self-host as a Python library only |
| Price | Self-hosted free within licence terms; API $1, $3 or $7.50 per 1,000 pages | Free, self-hosted (you pay for compute) |
| Output formats | Markdown, HTML, JSON, bounding boxes | Plain text, LaTeX, bounding boxes |
| Languages | 43-language average 77.8%; weak on Urdu (43.2%) and Sanskrit (51.1%) | Chinese and English only |
| Benchmark (self-reported) | 85.8 on olmOCR-bench (Chandra 2) | 90.57 on OmniDocBench v1.5 (OpenDoc-0.1B) |
| Known failure | Token-loop hallucinations reported on 5-6 of ~60 Russian/Kazakh scans; weak on old degraded scans | Document parsing needs a separate layout model, PP-DocLayoutV2; no hardware or throughput figures published |
Sources
What each one is
Chandra
Datalab's OCR model that keeps layout and handles forms, tables and handwriting.
Chandra detailsOpenOCR
Fudan's open toolkit for text detection and recognition, formulas, tables and document parsing.
OpenOCR 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/chandra","fallbacks": ["fudan/openocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Chandra alternativesOpenOCR alternativesBest OCR models in 2026