The development of AI bots that can conduct meaningful conversations like humans has been a major challenge for researchers. To create robots that can be used as intelligent assistants or emotional companions, the main focus of research is to make full use of massive unannotated data sets and improve open-domain dialogue systems. In this regard, Baidu’s PLATO-XL stands out with its 11 billion parameters and superior performance in Chinese and English conversation tasks. PLATO-2 won first place in five different dialogue tasks at DSTC-9, setting a record in the field of open-domain conversation. The upgrade to PLATO-XL now makes it the world’s largest Chinese and English dialog generation model, raising expectations for what hundred billion or even trillion parameter models might achieve when they become available.