Docling vs Typhoon OCR
Pick Docling to ingest born-digital PDFs, Office files and HTML into structured JSON for RAG, where it is one library for many formats under an MIT licence. Pick Typhoon OCR to read Thai and English scans, forms and books, where it is a dedicated 2B model with published Thai results. Docling is a document pipeline whose scan accuracy depends on the OCR engine you plug in. Typhoon OCR is an OCR model that only covers two languages.
Doclingibm/docling | 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 | 2.2 s / page | 1.8 s / page |
| Best for | Structured export for RAG | Thai documents |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | Latin scripts, CJK | 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 Docling if
- You have PDF, DOCX, PPTX, XLSX, HTML, EPUB, image or email inputs and want one API for all of them.
- You need lossless JSON that keeps layout, reading order and table structure, or Markdown, HTML, DocTags and plain text.
- You want an MIT-licensed library that runs on macOS, Linux and Windows, on CPU or GPU, with LangChain or LlamaIndex integrations.
Choose Typhoon OCR if
- Your documents are Thai or Thai-English mixed, such as government forms, financial reports or books.
- You want tables as HTML with merged cells, equations as LaTeX and checkboxes marked.
- You want either self-hosted weights (2B, or a 0.8B V2 variant) or the Typhoon API, which serves typhoon-ocr-v2.
Key differences
| Docling | Typhoon OCR | |
|---|---|---|
| What it is | Python library from IBM Research that converts many file formats into a structured DoclingDocument. | Thai and English document OCR model from Typhoon (SCB 10X). |
| How you run it | Local library. Plugs in OCR engines (EasyOCR, Tesseract, RapidOCR) and VLMs such as the 258M GraniteDocling. | Self-hosted 2B weights (0.8B V2 variant also exists), or the Typhoon API as typhoon-ocr-v2. |
| Licence and price | MIT, though individual models have their own licences. Free, you pay for compute. | Apache-2.0, subject to OpenTyphoon Terms and Conditions. Free self-hosted. API pricing: Not published. |
| Output formats | Markdown, HTML, JSON, DocTags, plain text. | Markdown, HTML, LaTeX. |
| Benchmarks | Not published. | V2 2B: olmOCR-bench 84.7, OmniDocBench v1.6 90.97, ThaiOCRBench 59.9. Scores are from the authors. |
| Languages | Not published. Scan accuracy depends on the OCR engine you plug in. | Only Thai and English evaluated. |
| Known failure | Version 2.143.1 drops words and numbers with full-page Tesseract OCR (issue #5254). Reading order can merge neighbouring columns (issue #5132). | V2 can loop until the token limit, mostly on dense multi-column pages. Key-information extraction is weakest, at 21.0 on ThaiOCRBench. |
Sources
- Docling GitHub README
- Supported formats
- Issue #5254 OCR regression
- Typhoon OCR V2 2B model card
- Typhoon OCR technical report
What each one is
Typhoon 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.
PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?
Use one, fall back to the other
{"model": "ibm/docling","fallbacks": ["typhoon/typhoon-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Docling vs Florence-2Docling vs GOT-OCR 2.0Tesseract vs Typhoon OCRMinerU vs Typhoon OCRLlamaParse vs DoclingPaddleOCR-VL vs DoclingMinerU vs DoclingDocling vs Tesseract
Docling alternativesTyphoon OCR alternativesBest OCR models in 2026