Qwen3.8-Flash-Next introduced as an early preview of Qwen4 architecture

This open-weight model with 125 billion parameters delivers a significant performance boost using only 6 billion active parameters, while also providing the first clues about Qwen4's future architecture.