Alibaba Announces Qwen 3.8-Flash-Next as Preview for Qwen 4

Alibaba's Qwen team will release Qwen 3.8-Flash-Next, a Mixture-of-Experts (MoE) model featuring 125 billion total parameters with 6 billion active parameters per token. Positioned as a preview for the Qwen 4 architecture, the model will be available on Hugging Face and ModelScope. The open-weight release aims to help developers prepare for the full Qwen series while reducing operational costs.