qwen3-235b-a22b is the flagship open-source MoE model in Alibaba Qwen3 family. With 235B total parameters and 22B active parameters, it supports a 128K context window. It natively enables switching between thinking and non-thinking modes, delivering open-source SOTA performance in math, coding and multilingual reasoning. It supports function calling and structured output under Apache2.0 license, suitable for complex reasoning, software development, long-document processing and Agent workflows.