Hunyuan OCR vs Qwen-VL
Pick Hunyuan OCR if you want a small self-hosted model (about 1B parameters) for photo and scene text, image translation, or low-resource and historical scripts, and you are outside the EU, UK and South Korea. Pick Qwen-VL if you need an Apache-2.0 licence, a hosted API option, or one model that also answers questions about the image. Qwen-VL can loop and repeat text on low-quality scans, so check its output if you transcribe at volume.
Hunyuan OCRtencent/hunyuan-ocr | Qwen-VLqwen/qwen-vl | |
|---|---|---|
| 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 | 2.0 s / page |
| Best for | Chinese documents and photos | Flexible extraction with prompts |
| Inputs | PDF, images | PDF, images |
| Languages | 100+ languages | 30+ languages |
| Max pages per request | 500 | 200 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| 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 Hunyuan OCR if
- It is about 1B parameters and still covers parsing, text spotting, information extraction and image translation in one model.
- It runs on a laptop, CPU or consumer GPU through the llama.cpp GGUF path.
- It exposes 12 task types via the vLLM client, including low-resource and ancient-script OCR.
Choose Qwen-VL if
- It is Apache-2.0 with no region exclusions, and it comes in sizes from 2B to 235B-A22B MoE.
- Alibaba Cloud also serves it as a hosted API, so you can skip running GPUs yourself.
- It reads text in 32 languages and outputs layout positions and a Qwen HTML document format, and it can also do reasoning and VQA on the same image.
Key differences
| Hunyuan OCR | Qwen-VL | |
|---|---|---|
| Licence | Tencent Hunyuan Community License. Excludes the EU, UK and South Korea. A separate licence is needed above 100M monthly active users. | Apache-2.0 (Qwen3-VL repository) |
| How you run it | Self-hosted only, via vLLM, Transformers or llama.cpp. | Self-hosted, or hosted API on Alibaba Cloud Model Studio. |
| Model size | About 1B parameters. | 2B, 4B, 8B, 32B dense; 30B-A3B and 235B-A22B MoE. |
| Hardware | NVIDIA GPU with CUDA 13 recommended for the vLLM setup. llama.cpp GGUF for CPU, consumer GPU or laptop. | FP8 checkpoints need an H100-class GPU and CUDA 12+. The 235B example uses 8-way tensor parallel on vLLM. |
| Output formats | Markdown, HTML, LaTeX, JSON, plain text, bounding boxes. | Plain text, HTML, bounding boxes, JSON, Markdown. |
| Benchmark | ParseBench-2 overall 67.11 and ParseBench mean 63.86 (v1.5). | Qwen 3 VL 8B scores 64.6 on olmOCR-bench. This is a different benchmark, so the numbers are not directly comparable. |
| Known failure | Archived 1.0 note said Transformers and vLLM outputs differed in accuracy. Needs CUDA 13 and vLLM 0.25.1+ for the unified environment. | Can loop and repeat the same text, worse with greedy decoding and on low-quality images. One report saw loops on 10-37% of OCR outputs after a transformers upgrade, versus 2-3% before. A Qwen2.5-VL-32B GGUF often got stuck looping on scanned documents. |
Sources
- HunyuanOCR model card
- Hunyuan Community License
- Qwen3-VL GitHub
- Issue #1611 infinite repetition
- olmo-eval issue #441
- Chandra README benchmark table
What each one is
Hunyuan OCR
Small end-to-end OCR model from Tencent covering parsing, extraction and photo translation.
Hunyuan OCR detailsQwen-VL
Alibaba's open-weight vision language models (Qwen2.5-VL, Qwen3-VL). Strong general OCR and grounding.
Qwen-VL 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": "tencent/hunyuan-ocr","fallbacks": ["qwen/qwen-vl"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Hunyuan OCR alternativesQwen-VL alternativesBest OCR models in 2026