chandra vs chunkr
Datalab's SOTA open OCR model: images/PDFs to structured HTML/Markdown/JSON with layout, tables, forms, checkboxes, handwriting and math, in 90+ languages. Local HF or vLLM inference. — versus — Document intelligence API in Rust: layout analysis, OCR with bounding boxes, and semantic chunking that turn PDFs, PPTs and Word docs into RAG-ready structured chunks.
Overlapping document-intelligence job at different layers: chunkr is the deployed pipeline API (layout + OCR + chunking); Chandra is the end-to-end OCR model you would slot into such a pipeline.
| chandra | chunkr | |
|---|---|---|
| Stars | 11k | 4.0k |
| Forks | 1.2k | 263 |
| Language | Python | Rust |
| License | Apache-2.0 | AGPL-3.0 |
| Last activity | 20 days ago | 3 months ago |
| Topics | ocr | ocr, rag |
| Curated connections | 4 | 5 |
chandra — the curator's take
Currently the strongest open OCR weights on the olmocr benchmark (85.8, above olmOCR 2 and dots.ocr), with handwriting, filled forms and checkboxes as the real differentiators — plus a serious self-built 90-language benchmark where it averages 72.7% vs Gemini 2.5 Flash's 60.8%. `pip install chandra-ocr`, `chandra_vllm`, done; ~2 pages/s real-world on an H100. The catch is licensing: code is Apache-2.0 but the WEIGHTS are OpenRAIL-M — free for research, personal use and sub-$2M startups, commercial self-hosting needs a Datalab license, and their paid API deliberately stays ahead of the open weights. Pick olmOCR for a fully permissive stack, MinerU when you want a whole parsing pipeline rather than the model itself.
chunkr — the curator's take
The RAG-ingestion specialist: where OCR tools stop at text, Chunkr does layout analysis and SEMANTIC chunking — the chunks arrive respecting document structure, with bounding boxes for citation-grounding. Self-hosted via Docker Compose. Read the split carefully though: the AGPL open-source version runs community models while the paid cloud runs proprietary ones — the README says plainly the accuracy differs; benchmark the OSS tier on YOUR documents before committing. Also ~3 months quiet at review time, and AGPL matters if you embed.