Alibaba releases Qwen3.8-Max, its largest 2.4T-parameter model

Qwen3.8-Max is a sparse MoE with 2.4T total parameters, 256K context, and native multimodal input, priced at $2/$6 per million tokens. It scores 56-58 on the Artificial Analysis Intelligence Index, trailing Kimi K3 by 1-4 points at lower cost. Open weights promised.