Qwen3-Coder

Qwen3-Coder is a new 480B MoE open model (35B active) by the Qwen team, built for agentic coding. It achieves SOTA results on benchmarks like SWE-bench, supports up to 1M context, and comes with an open-source CLI tool, Qwen Code.

Qwen3-Coder is a cutting-edge open-source model designed for agentic coding tasks, featuring a 480B Mixture-of-Experts architecture with 35B active parameters. It sets new benchmarks in coding performance, supporting up to 1M tokens of context and excelling in tasks like SWE-bench and tool-assisted coding. The model leverages advanced techniques such as large-scale reinforcement learning and synthetic data scaling to enhance accuracy and versatility.

Accompanying the model is Qwen Code, a CLI tool optimized for agentic coding, along with compatibility for Claude Code and Cline. Available via Alibaba Cloud’s API, Qwen3-Coder integrates seamlessly with developer workflows. Future developments aim to expand model variants and explore self-improving capabilities for complex software engineering tasks.