YAYI is a large language model developed by Zhongke Wenge (中科闻歌), fine-tuned from LLaMA 2 and BLOOM pretrained weights on million-level high-quality domain instruction data. The training data covers media publicity, public opinion analysis, public safety, financial risk control, and urban governance domains with hundreds of natural language tasks. The model offers enhanced Chinese capabilities, multi-turn dialogue, and some plugin abilities, with sizes including 7B and 13B parameters.