All tasks / Models
Chandra OCR 2
Datalab's open-weights OCR model that converts images and PDFs to Markdown, HTML or JSON with layout, tables, forms and handwriting.
- Datalab
- Open weights
- Released 18 Mar 2026
Where it ranks
- –Documents & OCRModels · Local · no result on this task’s benchmarks yet
Facts
- Vendor
- Datalab
- Family
- Chandra
- Released
- 18 Mar 2026
- Weights
- Open: download and run it yourself
- License
- Modified OpenRAIL-M (free for research, personal use and startups under $5M funding/revenue)
- Parameters
- 5.3B
- GPU memory to run locally
- ~13 GB (Estimate, no official figure: BF16 weights ≈ 5.3 B × 2 GB + ~1.5 GB runtime overhead.)
- Access
- Free (open weights)
- Last verified
- 18 Oct 2026
Price from datalab.to, 18 Oct 2026.
Benchmark results
| Benchmark | Result | Entry on the leaderboard | As of | Source |
|---|---|---|---|---|
| MDPBench (multilingual document parsing) | 79.7% | chandra-ocr-2Overall (public split, page-level mean of text, formula and table scores) from the board's leaderboard.json (updated 2026-10-04). | 4 Oct 2026 | huggingface.co |
| ParseBench (models) | 70.1% | Chandra-ocr-2Overall (mean of tables, charts, content faithfulness, semantic formatting, visual grounding) from leaderboard.csv in the ParseBench repo (last leaderboard commit 2026-10-09). | 9 Oct 2026 | github.com |
Where you can use it
- –MarkerParsers & APIs · Local
Earlier versions
- Chandra · 21 Oct 2025