amynicole – Alibaba’s Qwen team has introduced a powerful new open-source AI coding model, Qwen 3 Coder, aimed at developers and researchers. The model is designed with agentic capabilities. Enabling it to perform autonomous coding, use web browsers, and interact with development tools. According to Alibaba, this latest version is comparable in performance to Anthropic’s Claude Sonnet 4. The company released one variant so far: Qwen3-Coder-480B-A35B-Instruct, which stands as the most capable model in the lineup.
Read More : Trump’s AI Plan Challenges State Rules and Bias Concerns
The Qwen 3 Coder features a mixture-of-experts (MoE) architecture with a total of 480 billion parameters. Activating 35 billion parameters per operation. It also supports a native context length of 256,000 tokens, which can extend to one million tokens through extrapolation methods. These specifications allow for long and complex interactions without loss of coherence. The model is released under the Apache 2.0 license, permitting both commercial and academic use.
The team has made the model’s weights available through Hugging Face and GitHub, offering easy access for developers. To support its agentic coding capability, Alibaba also released Qwen Code, a command-line tool that integrates with the AI model. Based on Gemini Code, this tool supports writing, editing, deploying, and executing code inside an integrated development environment.
High Benchmark Scores and Multi-Platform Support Highlight Model’s Versatility
Alibaba claims that Qwen 3 Coder achieved state-of-the-art performance on the SWE-Bench Verified benchmark. Setting a new record among open-source coding models. This benchmark measures a model’s ability to solve real-world software engineering tasks. The Qwen team attributes its success to a scalable infrastructure setup on Alibaba Cloud. Which enabled parallel testing in 20,000 independent environments.
The Qwen Code tool enhances developer interaction with the model by supporting custom prompts and function calling protocols. These features enable seamless integration with software development environments, allowing the model to complete tasks with minimal human input. .
Read More : OpenAI Teams Up With Oracle to Expand Data Center Power
Developers interested in using the model with Claude Code can also do so. Although they must request an API key through the Alibaba Cloud Model Studio. This flexibility shows the model’s potential for inclusion in varied development workflows, from research environments to enterprise-level projects.
The introduction of Qwen 3 Coder marks a significant advancement in open-source AI coding tools. By combining large-scale capabilities with agentic functions and open licensing. Alibaba is positioning Qwen 3 Coder as a serious alternative to proprietary models in the autonomous coding space. Developers can now experiment with powerful AI tools that offer greater control, customization, and flexibility.
