Chemprop provides graph neural network models for molecular property prediction, specifically message passing neural networks that operate on molecular graph structures. It includes training pipelines, CLI tooling, and pretrained models for property prediction tasks common in cheminformatics and drug discovery.