Surya vs Typhoon OCR
Pick Surya if you need layout detection, reading order and bounding boxes across 90+ languages, run locally on a laptop or GPU. Pick Typhoon OCR if your documents are Thai or Thai-English, such as government forms, financial reports and books. The main split is language scope: Surya covers many languages with geometry, Typhoon OCR is tuned for Thai and English only.
Suryadatalab/surya | Typhoon OCRtyphoon/typhoon-ocr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 1.5 s / page | 1.8 s / page |
| Best for | Layout and line-level OCR | Thai documents |
| Inputs | PDF, images | PDF, images |
| Languages | 90+ languages | Thai, English |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | No | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | No |
| Forms | No | No |
| 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 Surya if
- You need line and block bounding boxes, layout labels and reading order from one model, across 90+ languages.
- You want to run on CPU or Apple Silicon through llama.cpp, with peak memory of about 4-6 GB on an M4 Max.
- You want throughput on a GPU: 4.52 pages/s at 192 DPI on one H100 with vLLM, from a 689M parameter model.
Choose Typhoon OCR if
- Your documents are Thai government forms, financial reports or Thai books, where V1.5 reached BLEU 0.870 and beat GPT-5 and Gemini 2.5 Pro on forms.
- You need tables as HTML with merged cells, equations as LaTeX and checkboxes marked.
- You want a choice of self-hosting or the Typhoon API (typhoon-ocr-v2), with Apache-2.0 weights subject to the OpenTyphoon terms.
Key differences
| Surya | Typhoon OCR | |
|---|---|---|
| Languages | 90+ languages; 91-language pass rate 74.9%, with 21 of 91 reaching 90% | Only Thai and English evaluated |
| Licence | Code Apache-2.0; weights under modified AI Pubs Open Rail-M, free for research, personal use and companies under $5M funding or revenue | Apache-2.0, subject to OpenTyphoon Terms and Conditions |
| How you run it | Self-host only, as a client against a vLLM or llama.cpp server you run | Self-host the weights or use the Typhoon API as typhoon-ocr-v2 |
| Model size | 689M | 2B (V2 and V1.5); 0.8B V2 variant |
| Output formats | JSON, HTML, LaTeX, bounding boxes, plain text | Markdown, HTML, LaTeX |
| olmOCR-bench | 83.7% (Surya 2.5) | 84.7 (V2 2B) |
| Known failure | Surya 2.5 table output lacks row and column geometry; multilingual tables score 58.1 | V2 can loop until the token limit on dense multi-column pages; key-information extraction scores 21.0 on ThaiOCRBench |
Sources
- Surya GitHub README
- Datalab pricing and license FAQ
- Typhoon OCR V2 2B model card
- Typhoon OCR technical report
What each one is
Surya
Open-source OCR toolkit with line detection, layout analysis, reading order and tables.
Surya detailsTyphoon OCR
Open vision model tuned for Thai and English documents, with Markdown output.
Typhoon 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/surya","fallbacks": ["typhoon/typhoon-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Surya alternativesTyphoon OCR alternativesBest OCR models in 2026