Chandra vs Nanonets-OCR
Pick Chandra for scanned, multilingual or table-heavy documents where you want Markdown, HTML or JSON with bounding boxes, and check the licence if your company is above $2M in funding or revenue. Pick Nanonets-OCR for printed business PDFs with signatures, watermarks and checkboxes, or for flowcharts you want as Mermaid, and you want a small 3B model with no hosted option to pay for. On olmOCR-bench Chandra 2 scores 85.8 against 69.5 for Nanonets-OCR2-3B, which is the main reason to choose Chandra for general OCR accuracy.
Chandradatalab/chandra | Nanonets-OCRnanonets/nanonets-ocr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.0 s / page | 1.8 s / page |
| Best for | Complex layouts and handwriting | Markdown with semantic tags |
| Inputs | PDF, images | PDF, images |
| Languages | 40+ languages | Multilingual |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | No |
| Forms | ✓ Yes | ✓ Yes |
| Multilingual | ✓ Yes | ✓ 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 Chandra if
- It scores 85.8 on olmOCR-bench (Chandra 2), against 69.5 for Nanonets-OCR2-3B.
- It averages 77.8% across 43 languages and returns JSON with bounding boxes and labels per block.
- You can self-host it or call the Datalab API at $1, $3 or $7.50 per 1,000 pages, with a free tier of $20 per month.
Choose Nanonets-OCR if
- It tags signatures, watermarks and image descriptions, and turns checkboxes into unicode symbols.
- It extracts flowcharts and org charts as Mermaid code and tells inline from display LaTeX equations.
- It is a 3B model fine-tuned from Qwen2.5-VL-3B, free to self-host, and scores 89.43 on DocVQA.
Key differences
| Chandra | Nanonets-OCR | |
|---|---|---|
| How you run it | Self-host with vLLM or Hugging Face, or use the hosted Datalab API | Self-host the weights from Hugging Face only |
| Licence | Code Apache-2.0; weights modified OpenRAIL-M, free under $2M funding or revenue, paid licence above that | Not stated on the Nanonets-OCR2-3B model card |
| Price | Free self-hosted; hosted API $1, $3 or $7.50 per 1,000 pages | Free self-hosted; you pay for compute |
| Output formats | Markdown, HTML, JSON, bounding boxes | Markdown, HTML, LaTeX, Mermaid |
| olmOCR-bench overall | 85.8 (Chandra 2) | 69.5 (OCR2-3B) |
| Known failure | Token-loop hallucinations reported on some Russian and Kazakh scans; weak on Urdu (43.2%) and Old Scans (51.1) | Not trained on handwriting; generation loops with CER up to 2100% in a handwriting test; hallucinates with prompts other than its fixed one |
| Hardware | One NVIDIA GPU via vLLM; benchmarked on an H100 80GB, with presets for A100, L40S and RTX 4090/5090 | Not published |
Sources
- Chandra GitHub README
- Datalab pricing
- GitHub issue #51 token loops
- Nanonets-OCR2-3B model card
- HF discussion fixed prompt
- Nanonets-OCR-s announcement
What each one is
Chandra
Datalab's OCR model that keeps layout and handles forms, tables and handwriting.
Chandra detailsNanonets-OCR
Open image-to-Markdown model that tags tables, equations, signatures, watermarks and checkboxes.
Nanonets-OCR 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": ["nanonets/nanonets-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Chandra alternativesNanonets-OCR alternativesBest OCR models in 2026