Abstract: On November 18, 2017, an MS 6.9 earthquake struck Nyingchi, Tibet Autonomous Region, China. The epicenter was located in the Yarlung Tsangpo Grand Canyon region in the eastern Himalayas of southeastern Tibet, triggering thousands of landslides. However, the susceptibility characteristics of these co-seismic landslides have not yet been comprehensively evaluated. Using this earthquake event as a case study, this study employed four machine learning models—Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), and Neural Network (NN)—to assess co-seismic landslide susceptibility on a GIS platform. Hyperparameters were optimized using the grid search method to improve model performance. The models were then evaluated and compared using ROC curves, AUC values, and multiple statistical metrics, and highly reliable co-seismic landslide susceptibility maps were generated. The results show that areas with high and very high susceptibility are mainly concentrated in steep valleys, along the seismogenic fault, and in areas near the epicenter and river channels. All four models exhibited strong predictive capability. Among them, the Random Forest model achieved the highest AUC value (0.92), followed by the Support Vector Machine model, whereas the Logistic Regression model yielded the lowest AUC value (0.86). Comparative analysis further indicates that the Random Forest and Support Vector Machine models performed best in predicting co-seismic landslide susceptibility. The susceptibility maps generated by these two models provide scientific support for regional co-seismic landslide risk prevention and hazard mitigation.