Chinese labs feeling the AGI: 5-10T parameter models, RSI Alibaba's latest announcements detail Qwen 4 currently in training, with Qwen 4.5 and Qwen 5 planned to reach 5-10 trillion parameters—up to four times larger than the existing 2.4 trillion parameter Qwen 3.8 Max. The Qwen team reports meaningful progress on recursive self-improvement, where models identify weaknesses, design experiments, and synthesize data to drive autonomous evolution with limited human involvement. Alibaba is advancing its full AI stack with the Zhenwu V900 chip for 500,000-card clusters, M890 supernodes handling over 2T parameter inference, and plans for over 20GW datacenter capacity by 2032, accelerating timelines to Q1 2027 production. Alibaba and Huawei are acting as if CXMT HBM will scale in 2027. Interesting bet by the best-informed customers.