Qwen 3

Getting Started

1. Try Qwen online at qwen.ai through the official chat interface for free.
2. Download model weights from Hugging Face or ModelScope in your preferred size and quantization.
3. Run Qwen locally with Ollama, vLLM, or the official transformers integration for fast inference.
4. Use the Qwen-Agent framework to build tool-using AI agents with built-in code execution capabilities.

Key Features

• Wide range of model sizes from 0.6B to 235B parameters including MoE variants for flexible deployment options.
• Best-in-class coding performance with strong results on HumanEval, MBPP, and real-world software engineering benchmarks.
• Advanced multilingual support covering 29+ languages with particularly strong Chinese and English capabilities.
• Vision-language models (Qwen-VL) for image understanding, OCR, and visual reasoning tasks.
• Apache 2.0 license for most model sizes, enabling fully unrestricted commercial and research use.
• Built-in tool use and function calling with native support for structured output and API integration.