Alibaba Qwen3.8 launches with open weights and Day-0 inference on eight platforms.

Alibaba released Qwen3.8 with open weights for both the 27B dense model and 2.4T-A95B MoE variant. The 27B model outperforms Qwen3.7-Plus and supports 262K native context extendable to 1M via YaRN, licensed under Apache 2.0. Eight inference providers — SGLang, vLLM, DigitalOcean, Modal, Nebius, Fireworks, and Together AI — offered Day-0 deployment with the full 1M context window.