Isothermal compression tests of as-cast 40Mn steel specimens were carried out at temperatures of 900-1100 °C and strain rates of 0.01-10 s -1 . Based on the obtained flow stress curves, four flow stress models, namely, strain-compensated, physically based, back propagation artificial neural network (BP-ANN), and modified BP-ANN based on a genetic algorithm (GA-BP-ANN), were established to predict the flow stress. The prediction performance of each model was assessed using the correlation coefficient (R) and average absolute relative error (AARE). The values of R from the strain-compensated, physically based, BP-ANN, and GA-BP-ANN models were 96.85, 98.53, 99.10, and 99.73%, respectively, while the values of AARE were 6.14, 5.95, 4.89, and 2.62%, respectively. The genetic algorithm provides the optimal initial weights and thresholds of the BP-ANN model. Thus, the GA-BP-ANN model has the highest accuracy and stability for predicting the flow stress. Integrated with finite element simulation, the four flow stress models were applied to predict the hot compression stroke responses of the 40Mn steel specimens. The results indicate that the GA-BP-ANN model has the best predictive ability and great potential for practical applications.