Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency

Alibaba's Qwen team open-sourced Qwen3.8-Flash-Next, a multimodal MoE model previewing the Qwen4 architecture. The 125B-parameter model with 6B activated per token trains at about 1/9 the cost of Qwen3.7-Plus, supports 262,144-token context, and is priced at $0.15 per million input tokens on QwenCloud.