Tongyi Qianwen Qwen3-Coder-Next Open Source: Small but Powerful!
Tongyi open-sources the Qwen3-Coder-Next coding agent model, using an MoE architecture with 80B total parameters, activating only 3B parameters per inference, significantly reducing computing power requirements. It achieved a problem-solving rate of over 70% on the SWE-Bench Verified benchmark, with performance close to larger-scale dense models.
Tongyi open-sources the Qwen3-Coder-Next coding agent model, using an MoE architecture with 80B total parameters, activating only 3B parameters per inference, significantly reducing computing power requirements. It achieved a problem-solving rate of over 70% on the SWE-Bench Verified benchmark, with performance close to larger-scale dense models.
February 4, 2026 — Introduction: Today, we're announcing Qwen3-Coder-Next, an open-weight language model designed specifically for coding agents and local development. The model is built on Qwen3-Next-80B-A3B-Base, adopting a new architecture with hybrid attention and MoE; through large-scale...
It features the following key enhancements:.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.