Alibaba Cloud · General LLM
Qwen
Alibaba's large language model family offering strong multilingual capabilities with particular strength in Chinese and English across diverse model sizes.
Overview
Qwen (Tongyi Qianwen) is Alibaba Cloud's family of large language models available in sizes from 0.5B to 110B parameters. Qwen 2 demonstrated competitive performance with leading Western models on both English and Chinese benchmarks, with particular strength in coding, mathematics, and multilingual tasks. The Qwen-VL multimodal variant adds vision understanding. The models are open-weight under permissive licenses, making them popular choices for applications targeting Chinese-speaking markets.
Parameters
0.5B to 110B (multiple variants)
Context Window
32K-128K tokens
Languages
29 languages (Chinese and English primary)
Modality
Text, vision (Qwen-VL), audio (Qwen-Audio)
License
Qwen License / Apache 2.0 (size dependent)
Capabilities
Bilingual Chinese-English text generation and understanding
Code generation and mathematical reasoning
Multimodal vision-language understanding (Qwen-VL)
Long-context processing up to 128K tokens
Tool use and function calling
Use Cases
Building AI applications for Chinese and bilingual markets
Deploying multilingual customer service systems
Creating coding assistants with strong Chinese documentation support
Developing multimodal AI applications with vision capabilities
Pros
- +Best-in-class Chinese language understanding and generation
- +Wide range of model sizes for different deployment needs
- +Multimodal variants cover vision and audio
- +Open weights with permissive licensing for most variants
Cons
- -English performance can trail specialized English models
- -Alibaba Cloud hosting may raise data sovereignty concerns
- -Less community tooling in Western developer ecosystems
- -Licensing terms vary across model sizes
Pricing
Open weights are free to download. Alibaba Cloud API pricing varies by model size. Qwen-Turbo: approximately $0.57/1M tokens.