Alibaba has released Qwen3.8-Max, its largest AI model to date, with 2.4 trillion parameters. This model utilizes a mixture-of-experts architecture, which allows for more efficient processing by only activating a portion of the model for each request. Meanwhile, DeepSeek's latest model, V4-Flash, has gained attention for its lower inference pricing compared to several competing systems. These developments indicate progress in China's efforts to improve the efficiency and affordability of AI models. This matters as it could lead to wider adoption and integration of AI technology in various industries.