New demands for passenger safety, vehicle performance and fuel economy have led to an increase in the use of advanced high strength steel. An increase in strength increases the springback behavior. Recently the development of high strength steel has rapidly advanced, requiring verification of earlier material models suitable for describing the elasticity and the plasticity behavior in sheet metal forming. The aim of this study is numerical implementation of a new constitutive model to improve the accuracy of finite element simulation of sheet metal forming processes for high strength steels, especially springback phenomenon. In order to achieve this, the use of a two-surface model was utilized. Experimental results showed that the behavior of the elastic deformation consists of two portion i.e., linear and non-linear parts. In the model, the linear elastic behavior separated from the non-linear behavior by using for an inner surface. By passing the first surface, non-linear elastic state is started. The stress-strain relationship in linear elastic regime defined by classical elastic modulus and non-linear regime is defined by using a variable elastic modulus. Plastic material behavior is also defined by the yield surface or outer surface. The stress-strain relationship is defined by using the elastic modulus at the end of the non-linear elastic regime. It should be noted that elasticity tensor remains constant during plastic deformation. The use of variable elastic modulus plays an important role in predicting springback high-strength steels. The model was implemented as a UMAT subroutine for Abaqus software and was then used a simulation of springback for the draw-bend process. The high strength steel used was the dual phase steel DP980. Finally, the results of prediction of springback for the model were compared with experimental results and QPE model. The results indicated a good correlation with experimental results and it was shown that the model could improve the prediction of springback compared to QPE model. Predicted springback has an error about 2-3% compared with experimental results. Keywords : High strength steels, Plasticity, Unloading behavior, Constitutive model, Springback.