Qwen-VL vs Typhoon OCR
Pick Typhoon OCR for Thai and English documents such as government forms, financial reports and books. It is a 2B model that scores 84.7 on olmOCR-bench, against 64.6 for Qwen 3 VL 8B in the independent Datalab table. Pick Qwen-VL when you need OCR in many languages, or one model that also does reasoning and VQA on photos and screenshots, and add loop checks for scans.
Qwen-VLqwen/qwen-vl | 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 | 2.0 s / page | 1.8 s / page |
| Best for | Flexible extraction with prompts | Thai documents |
| Inputs | PDF, images | PDF, images |
| Languages | 30+ languages | Thai, English |
| Max pages per request | 200 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | 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 Qwen-VL if
- It reads text in 32 languages, while Typhoon OCR has only been evaluated on Thai and English.
- It comes in sizes from 2B to 235B-A22B MoE, all under Apache-2.0, so you can match the model to your GPU.
- It handles OCR plus reasoning or VQA, and works on photos and screenshots with text, including low light, blur and tilted text per Qwen.
Choose Typhoon OCR if
- It is built for Thai and Thai-English documents, and V1.5 beat GPT-5 and Gemini 2.5 Pro on Thai government forms (BLEU 0.870).
- Its V2 2B scores 84.7 on olmOCR-bench, a figure the authors report, from a model far smaller than the Qwen sizes that have been benchmarked.
- It outputs tables as HTML with merged cells, equations as LaTeX, and marks checkboxes.
Key differences
| Qwen-VL | Typhoon OCR | |
|---|---|---|
| Languages | OCR in 32 languages | Thai and English only evaluated |
| Model sizes | 2B, 4B, 8B, 32B dense; 30B-A3B and 235B-A22B MoE | 2B (V2 and V1.5); 0.8B V2 variant |
| Licence | Apache-2.0 | Apache-2.0, subject to OpenTyphoon Terms and Conditions |
| Output formats | Plain text, HTML, bounding boxes, JSON, Markdown | Markdown, HTML, LaTeX |
| olmOCR-bench | 64.6 (Qwen 3 VL 8B, independent, Datalab table) | 84.7 (V2 2B, reported by the authors) |
| Hosted API pricing | Alibaba Cloud Model Studio, varies by model, not verified | Typhoon API (typhoon-ocr-v2), Not published |
| Known failure | Loops repeating the same text, worse with greedy decoding and on low-quality scans; GGUF builds often get stuck on scanned documents | V2 can loop until the token limit on dense multi-column pages; key-information extraction is weakest, 21.0 on ThaiOCRBench |
Sources
- Qwen3-VL GitHub
- Issue #1611 infinite repetition
- Chandra README benchmark table
- Typhoon OCR V2 2B model card
- Typhoon OCR technical report
What each one is
Qwen-VL
Alibaba's open-weight vision language models (Qwen2.5-VL, Qwen3-VL). Strong general OCR and grounding.
Qwen-VL 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": "qwen/qwen-vl","fallbacks": ["typhoon/typhoon-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Qwen-VL alternativesTyphoon OCR alternativesBest OCR models in 2026