Chandra vs GOT-OCR 2.0
Pick Chandra for scanned, multilingual documents and PDF to Markdown pipelines, where you want Markdown, HTML or JSON with bounding boxes and a hosted API option. Pick GOT-OCR 2.0 for formulas, sheet music, chemical structures and box-guided region OCR in English and Chinese, if you can self-host and accept non-commercial training data. The main split is scope: Chandra is built around page-level document conversion, GOT-OCR 2.0 around special content types.
Chandradatalab/chandra | GOT-OCR 2.0stepfun/got-ocr-2 | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 1 credit / page |
| Per 1,000 pages | $2.00 | $1.00 |
| Typical speed | 2.0 s / page | 1.5 s / page |
| Best for | Complex layouts and handwriting | Formatted text, formulas and charts |
| Inputs | PDF, images | images |
| Languages | 40+ languages | Chinese, English |
| Max pages per request | 500 | 100 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | No |
| Forms | ✓ Yes | No |
| Multilingual | ✓ Yes | No |
| 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
- You need a hosted option: the Datalab API costs $1, $3 or $7.50 per 1,000 pages, with a $20/month free tier.
- You process multilingual scans: Chandra 2 averages 77.8% across 43 languages, and the authors report 85.8 on olmOCR-bench.
- You want structured output: Markdown, HTML or JSON, plus layout chunks with bounding boxes and labels per block.
Choose GOT-OCR 2.0 if
- You read formulas, sheet music, chemical structures or charts, with LaTeX, TikZ, SMILES and kern output.
- You need region OCR guided by box coordinates or color, or multi-page OCR with batched inference in Hugging Face Transformers.
- You want a small 580M model that is free to self-host with no revenue cap in the code licence.
Key differences
| Chandra | GOT-OCR 2.0 | |
|---|---|---|
| How you run it | Self-host the open weights (vLLM or Hugging Face) or use the hosted Datalab API | Self-host only, via Hugging Face Transformers; CPU users are pointed to a llama.cpp port |
| Licence | Code Apache-2.0; weights modified OpenRAIL-M, free under $2M funding/revenue, paid licence above that | Code Apache-2.0; data CC BY-NC 4.0 |
| Price | Self-hosted: free within licence terms. API: $1, $3 or $7.50 per 1,000 pages | Free, self-hosted (you pay for compute) |
| Output formats | Markdown, HTML, JSON, bounding boxes | Plain text, Markdown, LaTeX, TikZ, SMILES, kern |
| Benchmarks | Chandra 2: 85.8 on olmOCR-bench; 77.8% on a 43-language average | Not published |
| Hardware | One NVIDIA GPU via vLLM; benchmarked on an H100 80GB, with presets for A100, L40S and RTX 4090/5090 | CUDA 11.8 and PyTorch 2.0.1 per the README |
| Known failure | Token-loop hallucinations on some Russian/Kazakh scans; weak on Urdu (43.2%) and old scans | Looping characters mid-recognition; wrong text on images with no text; format mode worse than plain OCR mode |
Sources
What each one is
Chandra
Datalab's OCR model that keeps layout and handles forms, tables and handwriting.
Chandra detailsGOT-OCR 2.0
Unified OCR model that reads plain text, tables, formulas, sheet music and charts.
GOT-OCR 2.0 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": ["stepfun/got-ocr-2"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Chandra alternativesGOT-OCR 2.0 alternativesBest OCR models in 2026