Alibaba’s Qwen team accelerated its release cadence in 2026 with the August 26 launch of Qwen3.8-Flash and open weights for Qwen3.8-Flash-Next, a 125B-parameter MoE model activating just 6B parameters per token that explicitly previews the hybrid GDN+QSA architecture planned for the Qwen4 family. The variant delivered strong coding and agent benchmarks while cutting training costs to roughly one-ninth of Qwen3.7-Plus, and September updates added omni-modal capabilities supporting 1M-token context across text, image, audio, and video. At the Apsara Conference, Alibaba confirmed Qwen4 is already in training alongside a custom Zhenwu V900 chip slated for 2027 mass production, positioning the lab to scale toward 5–10 trillion parameters in later series. Traders are watching for the first official Qwen4-Flash announcement or API pricing, as rapid iteration and open-weight previews continue to shape expectations around efficiency gains versus closed frontier competitors.