Qwen Code

Next-Generation AI Programming Assistant

2025 Most Powerful Open Source AI Programming Tool

Qwen Code

Next-Generation AI Programming Assistant

Powerful coding tool based on Qwen3-Coder-480B with ultra-long context and intelligent code generation

480B

Parameters

256K

Context Length

7.5T

Training Data

100%

Open Source

Core Features

Explore the revolutionary features of Qwen3-Coder, bringing unprecedented intelligent experience to modern software development

480B Ultra-Large Model

Based on 4800 billion parameter Qwen3-Coder, providing unprecedented code understanding and generation capabilities

256K Ultra-Long Context

Native support for 256K tokens, expandable to 1M tokens, handling large projects with ease

Intelligent Agent Interaction

Multi-turn reasoning and tool calling, supporting complex software engineering tasks and automation workflows

Reinforcement Learning Optimized

Trained with large-scale code reinforcement learning, significantly improving code execution success rate and quality

Quick Installation

Get started with Qwen Code in just a few simple steps

Install Node.js 20+

curl -qL https://www.npmjs.com/install.sh | sh

Install Qwen Code globally

npm i -g @qwen-code/qwen-code

Start coding

qwen -p 'create a React component'

Usage Examples

$ qwen -p 'create a React component'
# Generating code...
✓ Successfully generated React component
✓ Saved to ./components/MyComponent.tsx

Documentation & Resources

Access comprehensive guides and source code

View Documentation

Qwen Code is forked from Gemini CLI, so Gemini CLI documentation is applicable

GitHub Repository

Browse source code, report issues, and contribute to the project

Claude Code Integration

In addition to Qwen Code, you can now use Qwen3-Coder with Claude Code. Simply request an API key on Alibaba Cloud Model Studio platform and install Claude Code to start coding.

Install Claude Code globally

npm install -g @anthropic-ai/claude-code

We have provided two entrypoints for seamlessly experiencing coding with Qwen3-Coder.

Option 1: Claude Code Proxy API

Use environment variables to configure API access

Set environment variables:

export ANTHROPIC_BASE_URL=https://dashscope-intl.aliyuncs.com/api/v2/apps/claude-code-proxy

export ANTHROPIC_AUTH_TOKEN=your-dashscope-apikey

Then start coding:

claude

Option 2: Router Customization

Use claude-code-config for backend model customization

Install router packages:

npm install -g @musistudio/claude-code-router

npm install -g @dashscope-js/claude-code-config

Configure and run:

ccr-dashscope # Configure

ccr code # Start coding

Usage Examples

$ claude
# Starting Claude Code with Qwen3-Coder...
✓ Connected to Qwen3-Coder via DashScope
✓ Ready to assist with your coding tasks
❯ How can I help you code today?

Model Comparison

Understand the advantages of Qwen3-Coder compared to other mainstream AI programming tools

Feature	Qwen3-Coder	Claude Sonnet 4	GPT-4	ERNIE-4.5
Parameters	480B (35B activated)	Undisclosed	Undisclosed	300B
Context Length	256K (up to 1M)	200K+	128K	128K
Open Source	Yes	No	No	No

Open Source Advantage

As an open source model, Qwen3-Coder allows full customization and local deployment, ensuring data security and privacy protection while avoiding usage restrictions and cost issues of commercial models.

Frequently Asked Questions

Common questions and detailed answers to help you better use Qwen Code

What is Qwen Code?

Qwen Code is an official open-source command-line tool developed by Alibaba, based on the Qwen3-Coder large model. It supports Node.js 20+ and is designed for developers to easily integrate and extend, enabling code generation, review, and multi-turn interactive debugging.

How to install Qwen Code tool?

Installing Qwen Code requires two steps: 1. Install Node.js 20 or above; 2. Use the command 'npm i -g @qwen-code/qwen-code' to globally install the Qwen Code CLI tool.

What are the main features of Qwen Code?

Qwen Code's main features include: automatic code snippet generation via command line, code review and refactoring suggestions, multi-turn interactive code debugging, and seamless integration with mainstream tools like GitHub and VS Code.

What model is Qwen Code based on?

Qwen Code is based on the Qwen3-Coder-480B-A35B-Instruct large model, which is Alibaba's latest generation open-source Agentic coding model with 480 billion parameters (35B active parameters).

What development environments does Qwen Code support?

Qwen Code is built on Node.js and supports mainstream developer tools, compatible with community ecosystems. It can integrate with GitHub, VS Code, and other tools for quick deployment and team collaboration.

What is the context processing capability of Qwen Code?

Qwen Code is based on the Qwen3-Coder model that supports native 256K token ultra-long context, which can be extended to 1 million tokens with external extensions, suitable for handling large projects and complex multi-document tasks.

What use cases is Qwen Code suitable for?

Qwen Code is suitable for AI programming assistants, automated operations, intelligent code review, handling large code repositories and PRs, multi-turn reasoning and tool calling, especially for tasks requiring agentic code generation.

What advantages does Qwen Code have compared to other code generation tools?

Qwen Code is based on the open-source Qwen3-Coder model, excelling in code RL, long context, and multi-task agency. It supports local deployment and high customization, making it ideal for enterprise users requiring data security and customization.

What are the hardware requirements for using Qwen Code?

The Qwen Code tool itself has low hardware requirements, but locally deploying the complete Qwen3-Coder-480B-A35B-Instruct model requires extremely high VRAM and power supply demands. Regular consumer GPUs cannot run it, so cloud services are recommended.

Does Qwen Code support multi-turn interaction?

Yes, Qwen Code supports multi-turn interactive code debugging and generation. The underlying Qwen3-Coder model is trained with long-sequence RL (Agent RL), supporting complex software engineering tasks with planning, feedback, decision-making, and multi-turn interaction capabilities.

What is the quality of Qwen Code's training data?

Qwen Code is based on the Qwen3-Coder model trained on 7.5 trillion tokens of high-quality data, with 70% being high-quality code. The training dataset was cleaned and rewritten using Qwen2.5-Coder, significantly improving the training set quality.

How to get more resources and support for Qwen Code?

You can visit the official blog to access more Qwen Code related resources and download methods. Qwen Code also has an active open-source community that supports developer communication and technical support.

Developer Reviews

Hear what the developer community really thinks about Qwen Code

"Finally waited for the successor to qwen2.5-coder 32b! Code generation quality has improved significantly."

DevMaster2024

"In terms of code RL and multi-turn interaction, Qwen3-Coder is the new benchmark in the open source field."

AIEnthusiast

"Claude has been performing poorly recently, Qwen3-Coder has become the ideal choice for local deployment."

TechReviewer

Average Rating: 4.8/5.0