CogView2 is a large-scale generative model for text-to-image synthesis in both Chinese and English. The model uses a hierarchical transformer architecture with 6B-9B-9B parameters and includes LoPAR for accelerated generation and CogLM for bidirectional completion. This implementation provides pretrained models for text-to-image generation and text-guided completion, deployed via Huggingface Spaces and Replicate web demos.