At the 2026 Yunqi Conference, Alibaba showcased the cutting-edge progress of its large-scale models. Regarding large language models, Qwen3.8-Max has demonstrated outstanding performance in programming and office scenarios, securing two international evaluation championships. The training process for Qwen4, which is built upon a next-generation architecture, has already kicked off. Future iterations are expected to scale up parameters to the range of 5 to 10 trillion. In the realm of multimodal models, a variety of multimodal, voice, world, and video generation models from the Tongyi Qianwen series have exhibited remarkable capabilities. Some of these models have earned a place at the international vanguard or even claimed the top spots. Additionally, Alibaba is currently training its next-generation video model, which is poised to drive AI's evolution from a simple tool to a source of creative intelligence.