Qwen3-Coder 480B-A35B Instruct

Qwen3-Coder-480B-A35B-Instruct is a causal language model built on a Mixture-of-Experts architecture, designed specifically to serve as an autonomous coding agent. With 480 billion total parameters and 35 billion active parameters across 160 experts, the model balances high-level reasoning with computational efficiency. It is engineered to handle complex workflows, featuring native support for a 256,000-token context window that can be extended to 1 million tokens using Yarn, making it particularly effective for understanding and navigating entire code repositories.

The model underwent comprehensive pre-training and post-training phases to refine its performance in agentic coding, browser-use, and tool-use scenarios. By utilizing a specialized function-calling format, it integrates seamlessly with developer tools and environments, including the Qwen Code CLI. Its design focuses on delivering state-of-the-art results on rigorous benchmarks like SWE-bench-Verified and WebArena, positioning it as a robust, open-weight solution for developers who require an AI capable of managing multi-turn, interactive, and autonomous software engineering tasks.