Glint360K is the largest and cleanest face recognition dataset to date, containing 17,091,657 images of 360,232 people. The dataset can be used to train a baseline model using the Partial FC training strategy. Partial FC is a distributed deep learning training framework for face recognition that is designed to advance large-scale classification tasks (e.g., 10 million or 100 million identities). The framework is faster than parallel solutions of the model and guarantees performance.