CLIP (Contrastive Language–Image Pre-training) is a vision-language model developed by OpenAI that learns to connect images with natural language descriptions. In this post, I will review its architecture and training approach, and collect resources for understanding and working with the model.