In order to predict the high-temperature deformation behavior of hypereutectoid steel, the hot compression tests were conducted in the strain rate range of (0.001∼1) s −1 and the deformation temperature range of (950∼1100) • C. The experimental data were employed to develop the Arrhenius constitutive model and BP neural network model, and their predictability for high temperature flow stress of hypereutectoid steel was further evaluated. Comparatively, a higher correlation coefficient (R) can be obtained for the BP neural network model compared with the Arrhenius constitutive equations. And the relative error within ±1% was more than 68.75% for the BP neural network model, while only 18.75% for the constitutive equations. The BP neural network model is considered more efficient and accurate to predict the hot deformation behavior than the Arrhenius constitutive equations. Moreover, the well-trained BP neural network model is employed to predict the flow stress varying with the deformation temperature and strain rate. The flow stress decreases with the increasing deformation temperature and decreasing strain rate, which is in accordance with the experimental evaluation. The results indicate that BP neural network model is an efficient tool for modelling and predicting the flow behavior of hypereutectoid steels in high temperature applications.INDEX TERMS Hypereutectoid steel, hot deformation behavior, Arrhenius equations, BP neural network models.