Granite Docling vs Typhoon OCR
Pick Granite Docling for English PDFs with tables, code and equations, especially inside a Docling pipeline on CPU or a small GPU. Pick Typhoon OCR for Thai or Thai-English documents such as government forms, financial reports and books. The main split is language: Granite Docling is English-first and marks Japanese, Arabic and Chinese as experimental, while Typhoon OCR is built and evaluated for Thai and English.
Granite Doclingibm/granite-docling | Typhoon OCRtyphoon/typhoon-ocr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 1 credit / page | 2 credits / page |
| Per 1,000 pages | $1.00 | $2.00 |
| Typical speed | 1.5 s / page | 1.8 s / page |
| Best for | Small-model document conversion | Thai documents |
| Inputs | PDF, images | PDF, images |
| Languages | English, experimental CJK and Arabic | Thai, English |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | No | ✓ 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 Granite Docling if
- It has only 258M parameters, so it runs cheaply on CPU or a small GPU, and it plugs into the Docling library.
- Its authors report table structure TEDS of 0.97 on FinTabNet, code recognition F1 of 0.988 and equation recognition F1 of 0.968.
- It outputs DocTags, Markdown, HTML, LaTeX and bounding boxes, and it is Apache-2.0 with no extra terms.
Choose Typhoon OCR if
- It targets Thai: V1.5 scored BLEU 0.870 on Thai government forms, ahead of GPT-5 and Gemini 2.5 Pro, and V2 2B scores 81.3 on MDPBench Thai pages versus 61.3 for its base model.
- You can self-host the weights or call the Typhoon API as typhoon-ocr-v2, so you can start hosted and move to your own servers later.
- It writes tables as HTML with merged cells, equations as LaTeX, and marks checkboxes, which suits forms and financial reports.
Key differences
| Granite Docling | Typhoon OCR | |
|---|---|---|
| Language focus | English primary; Japanese, Arabic and Chinese are experimental | Thai and English; no other languages evaluated |
| How you run it | Self-host only, typically inside Docling | Self-host the weights or use the Typhoon API (typhoon-ocr-v2) |
| Licence | Apache-2.0 | Apache-2.0, subject to OpenTyphoon Terms and Conditions |
| Model size | 258M parameters | 2B (V2 and V1.5); 0.8B V2 variant |
| Output formats | DocTags, Markdown, HTML, LaTeX, bounding boxes | Markdown, HTML, LaTeX |
| Published benchmarks | Table TEDS 0.97 (FinTabNet), full-page OCR F1 0.84, OCRBench 500 | olmOCR-bench 84.7, OmniDocBench v1.6 90.97, ThaiOCRBench 59.9 (V2 2B) |
| Known failure | Repetitive loops on page headings and invoices; may output only exclamation marks on GPUs without bfloat16 such as T4 unless run in float32 | V2 can loop until the token limit on dense multi-column pages; weakest at key-information extraction (21.0 on ThaiOCRBench) |
Sources
- Granite-Docling model card
- HF discussion #24 repetition
- Typhoon OCR V2 2B model card
- Typhoon OCR technical report
What each one is
Granite Docling
Compact IBM vision model that converts pages to DocTags for the Docling pipeline.
Granite Docling 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": "ibm/granite-docling","fallbacks": ["typhoon/typhoon-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Granite Docling alternativesTyphoon OCR alternativesBest OCR models in 2026