Chatting with Tongyi Qianwen in DingTalk to test Alibaba's large language model, three classic questions were asked. 1: The image understanding effect is very impressive, accurately describing scenes, objects, and text on the screen; 2: Text-to-image still has the classic finger problem; 3: Logical reasoning questions are clear and accurate, with better performance on the same questions compared to Wenxin Yiyan and kimi (no screenshots here). Embedding a chatbot in a messaging tool feels very natural, whereas embedding bots in other apps generally has lower penetration because users don't know how to ask questions—they first try to solve problems through the app's native functionality. In messaging tools, users seem more inclined to use conversation to resolve issues.$Alibaba(BABA.US)