Alibaba to Release Qwen 3.8-Flash-Next, Previewing Qwen 4 Architecture

Alibaba's Qwen team will release Qwen 3.8-Flash-Next on Wednesday, a mixture-of-experts (MoE) model with 125 billion total parameters, activating only 6 billion parameters per token. The team positions it as a preview of the next-generation Qwen 4 architecture, rather than a final flagship model. The model is described as multimodal and built on the upcoming Qwen 4 architecture. The team stated that releasing this early version ahead of time is intended to allow developers to prepare for the subsequent full Qwen series. Model weights will be hosted on Hugging Face and ModelScope platforms. As of now, Qwen has not published benchmark results for the model, nor comparative data against its own Qwen 3 series or overseas competitors, and the specific figures of 125 billion and 6 billion parameters have not been officially confirmed. As an open-weight model, developers can download, fine-tune, and run it for free without sending data to a closed API, helping to reduce the cost of hosting models.