Qwen and Qwen-VL
These are ecosystem siblings: Qwen-VL extends the base Qwen language model with multimodal vision-language capabilities, allowing users to process both text and images while building on the same underlying architecture and training approach.
About Qwen
QwenLM/Qwen
The official repo of Qwen (通义千问) chat & pretrained large language model proposed by Alibaba Cloud.
Supports multiple quantization schemes (Int4/Int8, GPTQ) and context lengths up to 32K tokens, with chat variants fine-tuned via SFT and RLHF for tool use, code generation, and agent capabilities. Models range from 1.8B to 72B parameters across base and instruction-tuned variants, with Q-LoRA fine-tuning support. Available on Hugging Face and ModelScope with deployment examples via vLLM and FastChat, plus OpenAI-compatible API integration.
About Qwen-VL
QwenLM/Qwen-VL
The official repo of Qwen-VL (通义千问-VL) chat & pretrained large vision language model proposed by Alibaba Cloud.
Supports ultra-high resolution images up to millions of pixels with extreme aspect ratios, enabling detailed text recognition and document analysis. Built on a vision-language architecture optimized for reasoning tasks, the model family includes quantized Int4 variants and API-accessible Plus/Max versions. Integrates with HuggingFace, ModelScope, and Alibaba's DashScope API, with fine-tuning support via full-parameter, LoRA, and Q-LoRA methods.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work