Abstract: To investigate the high-temperature rheological behavior and constitutive relationship of 5456 aluminum alloy, isothermal compression experiments were conducted under the temperature 573-773 K and the strain rate 0.01-10 s-1, and the rheological curves of material were obtained. Then, based on the experimental datas, Arrhenius (AH) with strain compensation, Hensel-Spittel (HS) and Back-Propagation (BP) artificial neural network models were established. Finally, for the shortcomings of existing models, a new constitutive model was proposed based on the linear combinations of logarithmic stress, temperature, logarithmic temperature, logarithmic strain rate and quadratic logarithmic strain rate. The results indicate that with the increasing of strain, the stress of  5456 aluminum alloy shows a trend of first rapidly increasing, then slowly decreasing and finally stabilizing. In terms of prediction accuracy, BP neural network model performs the best, followed by the new model, then AH model, and finally HS model. However, the new model has a concise parameter form and good interpretability, making it suitable for the numerical simulation with high-precision reguirement.