The Alibaba QianWen team has officially launched Qwen3-Coder-Next, an open-weight language model tailored for coding agents and local development environments. Based on the Qwen3-Next-80B-A3B-Base framework, this model incorporates a cutting-edge architecture and employs specialized training techniques to deliver robust programming and agent functionalities, all while minimizing inference expenses. Qwen3-Coder-Next places a strong emphasis on enriching agent training signals, demonstrating exceptional performance across various programming agent benchmark tests and striking an optimal balance between efficiency and effectiveness. Despite its promising potential for diverse applications, there remains ample opportunity for refinement. Looking ahead, the team is committed to bolstering its reasoning and decision-making prowess, accommodating a broader spectrum of task types, and facilitating swift iterative enhancements. The model's open-source repositories can be accessed on ModelScope (https://www.modelscope.cn/collections/Qwen/Qwen3-Coder-Next) and Hugging Face (https://huggingface.co/collections/Qwen/qwen3-coder-next).