Qianfan OCR vs Typhoon OCR
Pick Qianfan OCR for self-hosted parsing across many languages, with tables, formulas, handwriting and field extraction. Pick Typhoon OCR for Thai and English documents such as government forms, financial reports and books, where it is small (2B) and tuned for Thai. The main split is language coverage: Qianfan OCR covers 192 languages, while Typhoon OCR was only evaluated on Thai and English.
Qianfan OCRbaidu/qianfan-ocr | Typhoon OCRtyphoon/typhoon-ocr | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.0 s / page | 1.8 s / page |
| Best for | Chinese documents | Thai documents |
| Inputs | PDF, images | PDF, images |
| Languages | Chinese, English and more | Thai, English |
| Max pages per request | 200 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | No |
| Bounding boxes | No | 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 Qianfan OCR if
- You need more than Thai and English: it covers 192 languages, plus handwriting and scene text.
- You need key information extraction, tables and formulas, with output as Markdown, JSON, HTML, LaTeX or bounding boxes.
- You want a clear Apache-2.0 licence for commercial use, with no extra terms listed.
Choose Typhoon OCR if
- Your documents are Thai or mixed Thai and English, such as government forms, financial reports and books.
- You want a smaller model: V2 is 2B, with a 0.8B variant, and it scores 84.7 on olmOCR-bench.
- You need tables as HTML with merged cells, equations as LaTeX and checkboxes marked.
Key differences
| Qianfan OCR | Typhoon OCR | |
|---|---|---|
| Languages | 192 languages, plus handwriting and scene text | Only Thai and English evaluated |
| Model size | 4B (Qwen3-4B language model) | 2B for V2 and V1.5, 0.8B V2 variant |
| Licence | Apache-2.0 | Apache-2.0, subject to OpenTyphoon Terms and Conditions |
| How you run it | Self-host the weights (benchmarked on one A100 with vLLM) or use the Baidu Qianfan platform | Self-host the weights or use the Typhoon API as typhoon-ocr-v2 |
| Output formats | Markdown, JSON, HTML, LaTeX, bounding boxes | Markdown, HTML, LaTeX |
| olmOCR-bench | 79.8 | 84.7 (V2 2B) |
| Known failure | Thinking mode can emit up to 16,384 tokens, adding latency; 1.024 pages per second on an A100 even with W8A8 | V2 can loop until the token limit on dense multi-column pages; key information extraction is its weakest task at 21.0 on ThaiOCRBench |
Sources
What each one is
Qianfan OCR
Baidu's OCR-tuned vision model. Strong on Chinese documents and tables.
Qianfan OCR 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": "baidu/qianfan-ocr","fallbacks": ["typhoon/typhoon-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Qianfan OCR alternativesTyphoon OCR alternativesBest OCR models in 2026