The proposed CLIP Surgery is a method that enables surgery-like modifica-tions for the inference architecture and features, for better explainability and enhancement in multiple open-vocabulary tasks and demonstrates remarkable improvements in open-vocabulary segmentation and multi-label improvements.