Chandra
OCR model that converts PDFs and images into HTML, Markdown, or JSON with layout, tables, and handwriting.
About
Datalab, the team behind Marker and Surya, trained Chandra to turn images and PDFs into structured HTML, Markdown, or JSON while preserving layout information. The vision language model handles complex tables, forms, handwriting, and mathematical notation, extracts embedded images with captions, and reads more than 90 languages, averaging 77.8 percent across 43 languages on the project's multilingual benchmark. Installation is pip install chandra-ocr with a choice of vLLM or Hugging Face inference backends, and throughput reaches about 1.44 pages per second on an NVIDIA H100; an NVIDIA GPU is effectively required for practical use. The code is Apache-2.0, but the weights carry a modified OpenRAIL-M license: free for research, personal use, and companies under 2 million dollars in funding or revenue, with commercial licensing sold by Datalab beyond that threshold. The repository has about 11,800 stars, and the Chandra 2 release of March 2026 scores 85.8 on the olmOCR benchmark. Document pipelines adopt it where handwriting, forms, and multilingual scans defeat conventional OCR engines.
Should you use Chandra?
Pick it when
Choose Chandra when handwriting, filled forms, math, or scans in many languages defeat classic OCR, you have NVIDIA GPU capacity, and your organization is under the 2 million dollar threshold or will buy a Datalab license.
Look elsewhere when
Skip it if you are a larger company that wants free commercial use: dots.ocr and DeepSeek-OCR ship MIT weights. Without an NVIDIA GPU, a CPU-capable pipeline such as Docling or Tesseract is the practical route.
Alternatives to Chandra
- Surya OCR
Same team and a similar weights license with a higher 5 million dollar cap, on a 650M model that also runs on CPU via llama.cpp, but it returns OCR geometry as JSON rather than formatted pages.
- dots.ocr
MIT weights with no revenue cap and strong low-resource script coverage, though handwriting and forms are not its stated focus the way they are for Chandra.
- olmOCR
Apache-2.0 code and weights for unrestricted commercial use and also handles handwriting, but outputs plain text and Markdown rather than HTML or JSON with layout.
- Marker
Datalab's PDF converter with a CPU fast mode and a weights license free up to 5 million dollars, but handwriting and filled forms are Chandra's stated focus, not Marker's.
Reviews (0)
Leave a Review
No reviews yet. Be the first to review!
Details
- Category
- OCR & Document Processing
- Price
- Freemium
- Platform
- Local/Desktop
- Difficulty
- Intermediate (3/5)
- License
- OpenRAIL-M
- Added
- Jul 29, 2026