Detailed descriptions of the mechanical properties of materials under different strain rates are necessary in order to ensure the reliability of finite element analysis under high-speed deformation. In this study, different types of alloys are examined in terms of their dynamic hardening characteristics. In addition to the Johnson–Cook model, Zerilli-Armstrong model, modified Khan-Huang model, modified Johnson–Cook model, and modified Lim–Huh model, the new calibration method and stepwise calibration method were used to study and evaluate a number of well-known dynamic hardening models: Johnson–Cook model, Zerilli-Armstrong model, modified Johnson–Cook model, modified Lim–Huh model. The feed-forward back-propagation neural networks are trained and validated to predict flow behaviors at different strain rates. Calculated models predict the hardening curve at different strain rates and then compare them with experimental results. Following analysis of the structural properties and calibration results of different models, it was found that the new global calibration method was more accurate than the stepwise calibration method when predicting dynamic hardening characteristics. It can greatly improve fit accuracy for alloys whose flow stress doesn’t increase monotonically with strain rate. Furthermore, the neural network is found to be the most accurate method to predict a material’s rate-dependent behavior.