Docling vs Chandra
Pick Docling to ingest born-digital PDFs, Office files, HTML and emails into structured JSON for RAG, under an MIT licence with no GPU required. Pick Chandra to read scanned and multilingual pages, or pages with tables, forms and math, and return Markdown, HTML or JSON with bounding boxes. The main split: Docling is a format-converting library whose scan accuracy depends on the OCR engine you plug in, while Chandra is an OCR model built for the page image itself.
Doclingibm/docling | Chandradatalab/chandra | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.2 s / page | 2.0 s / page |
| Best for | Structured export for RAG | Complex layouts and handwriting |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | Latin scripts, CJK | 40+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| 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 Docling if
- You need one API for PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails and XBRL.
- You want lossless JSON that keeps layout, reading order and table structure for LangChain or LlamaIndex pipelines.
- You want an MIT licence and CPU-only runs on macOS, Linux or Windows, with no revenue limit on commercial use.
Choose Chandra if
- Your documents are scans in many languages: Chandra 2 averages 77.8% across 43 languages in the authors' benchmark.
- You need per-block bounding boxes and labels, or pages full of tables, forms and math, converted to Markdown, HTML or JSON.
- You would rather call a hosted API than run models yourself: Datalab charges $1, $3 or $7.50 per 1,000 pages, and the free tier gives $20 a month.
Key differences
| Docling | Chandra | |
|---|---|---|
| Licence | MIT (individual models have their own licences) | Code Apache-2.0; weights modified OpenRAIL-M, free for research, personal use and companies under $2M funding or revenue |
| How you run it | Python library, self-hosted, CPU works and GPU speeds up model stages | Self-host on a single NVIDIA GPU via vLLM, or use the hosted Datalab API |
| Price | Free, you pay for compute | Self-hosted free within licence terms; hosted API $1 (fast), $3 (balanced) or $7.50 (accurate) per 1,000 pages |
| Input formats | PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails, XBRL | Page images and PDFs to Markdown, HTML or JSON |
| Output | Markdown, HTML, JSON, DocTags, plain text | Markdown, HTML, JSON, bounding boxes |
| Benchmark | Not published | olmOCR-bench overall 85.8 (Chandra 2, authors' table) |
| Known failure | Version 2.143.1 drops words and numbers with full-page Tesseract OCR (issue #5254); reading order can merge neighbouring columns (issue #5132) | Token-loop hallucinations on some Russian and Kazakh scans (issue #51); weak on Urdu (43.2%) and old scans (51.1 on olmOCR-bench) |
Sources
- Docling GitHub README
- Supported formats
- Issue #5254 OCR regression
- Chandra GitHub README
- Datalab pricing
- GitHub issue #51 token loops
What each one is
Chandra
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": "ibm/docling","fallbacks": ["datalab/chandra"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Docling alternativesChandra alternativesBest OCR models in 2026