This article proposes a method for predicting comprehensive tooth information from intra-oral scans, including tooth instance segmentation, tooth FDI numbering, and tooth landmark detection and classification. The method is implemented using two successive stages. The first stage for end-to-end tooth instance segmentation and labeling uses a large context optimized for tooth labeling. The second stage for landmark detection and classification processes high-resolution tooth crops to predict a more precise tooth segmentation and independently proposes landmarks of each landmark class. Our method achieved the first place in the Final Test Phase of the 3DTeethLand challenge and shows promise for application in dentistry and orthodontics to automate currently time-consuming clinical tasks.