Creep rupture life is an important target in the design of Ni-based single crystal superalloys. Predicting the creep rupture life of potential new alloys accurately can narrow the composition searching space and shorten the alloy development cycle. In this work, we established an end-to-end deep learning model named Superalloy Transformer-based Net for Creep (SaTNC), to predict the creep rupture life of Ni-based single crystal superalloys with high reliability. The relationship among chemical composition, service environment, microstructure characteristics, and the creep rupture life is constructed. In SaTNC, the prior knowledge is integrated into the representation of chemical composition information, including the features fusion of the element intrinsic properties and content, as well as the chemical information sharing between the elements in the alloy system by introducing the Self-attention mechanism in Transformer. The predictions of the SaTNC model are relatively accurate in the composition, temperature, and stress sensitivity of the alloy. The SaTNC model also shows better generalization performance compared with other advanced machine learning models. Besides, the SaTNC model can provide interpretable information and help to propose material science insights about the key alloying elements affecting creep rupture life as well as the interactions among these elements. Moreover, the elemental embeddings extracted from SaTNC encode the knowledge of the partitioning behaviors in γ/γ' and intrinsic properties of alloying elements, which are expected to improve the prediction performance for the task with small samples by combining with transfer learning. The constructed SaTNC model based on deep learning can guide the composition design and life assessment of modern Ni-based single crystal superalloys. More generally, it can be further applied in the design of other alloy materials.