SchNet is a deep learning architecture designed to model atomistic systems, particularly molecules and materials. It uses continuous-filter convolutional layers to learn representations of atom types and their interactions, capturing fundamental symmetries such as rotational and translational invariance. The architecture is trained to predict a range of properties across chemical space, including potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations. SchNet demonstrates accurate predictions for various properties of small molecules and materials, outperforming other models in terms of accuracy and efficiency. It also provides insights into the learned representations through local chemical potentials, revealing fundamental chemical knowledge from the periodic table. Additionally, SchNet enables path-integral molecular dynamics simulations with significantly reduced computational costs, making it a valuable tool for exploring complex chemical and material systems.SchNet is a deep learning architecture designed to model atomistic systems, particularly molecules and materials. It uses continuous-filter convolutional layers to learn representations of atom types and their interactions, capturing fundamental symmetries such as rotational and translational invariance. The architecture is trained to predict a range of properties across chemical space, including potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations. SchNet demonstrates accurate predictions for various properties of small molecules and materials, outperforming other models in terms of accuracy and efficiency. It also provides insights into the learned representations through local chemical potentials, revealing fundamental chemical knowledge from the periodic table. Additionally, SchNet enables path-integral molecular dynamics simulations with significantly reduced computational costs, making it a valuable tool for exploring complex chemical and material systems.