Surya vs GOT-OCR 2.0
Pick Surya when you need layout analysis, reading order and bounding boxes for 90+ languages, or want OCR on a laptop via CPU or Apple Silicon. Pick GOT-OCR 2.0 for formulas, sheet music, chemical structures or OCR of a specific region, mainly on English and Chinese documents. Surya publishes benchmark scores and GOT-OCR 2.0 does not, so Surya is easier to size up before you commit.
Suryadatalab/surya | 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 | 1.5 s / page | 1.5 s / page |
| Best for | Layout and line-level OCR | Formatted text, formulas and charts |
| Inputs | PDF, images | images |
| Languages | 90+ languages | Chinese, English |
| Max pages per request | 500 | 100 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | No | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | No |
| Forms | No | 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 Surya if
- It returns layout labels, reading order, OCR, HTML tables and KaTeX math with per-block bounding boxes from one model.
- It runs on NVIDIA GPU via vLLM or on CPU and Apple Silicon via llama.cpp, at about 4.3-5.9 GB peak memory on an M4 Max.
- It scores 83.7% on olmOCR-bench (Surya 2.5) and reached 4.52 pages/s at 192 DPI on one H100 with vLLM.
Choose GOT-OCR 2.0 if
- It reads formulas, sheet music, chemical structures (SMILES), charts and geometry, and emits LaTeX, TikZ and kern.
- It supports region OCR guided by box coordinates or color, plus multi-page OCR and dynamic resolution.
- It is merged into Hugging Face Transformers with batched inference, so setup is a Transformers load rather than a separate inference server.
Key differences
| Surya | GOT-OCR 2.0 | |
|---|---|---|
| Licence | Code Apache-2.0; weights under a modified AI Pubs Open Rail-M licence (free for research, personal use and companies under $5M funding/revenue) | Code Apache-2.0; data CC BY-NC 4.0 |
| Cost | Free self-hosted; commercial licence needed above $5M revenue or funding | Free self-hosted; you pay for compute |
| How you run it | Lightweight client against a separately running vLLM or llama.cpp server | Hugging Face Transformers; README lists CUDA 11.8 and PyTorch 2.0.1, with a llama.cpp port for CPU |
| Size | 689M parameters | 580M parameters |
| Output formats | JSON, HTML, LaTeX, bounding boxes, plain text | Plain text, Markdown, LaTeX, TikZ, SMILES, kern |
| Published benchmarks | olmOCR-bench 83.7%, OmniParseBench 85.9, 91-language pass rate 74.9% (author-reported) | Not published |
| Known failure | Surya 2.5 table output lacks row/column geometry; multilingual tables score 58.1; not for photos or natural scenes | Reported issues: looping characters mid-recognition, wrong text on blank images, and format mode worse than plain OCR mode |
Sources
- Surya GitHub README
- Datalab pricing and license FAQ
- GOT-OCR2.0 GitHub
- GOT-OCR 2.0 paper
- GitHub issue index
What each one is
Surya
Open-source OCR toolkit with line detection, layout analysis, reading order and tables.
Surya 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/surya","fallbacks": ["stepfun/got-ocr-2"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Surya alternativesGOT-OCR 2.0 alternativesBest OCR models in 2026