This paper presents a new dual mode nonlinear model predictive controller (NMPC) that is based on the combination of the finite horizon NMPC with the infinite horizon predictive controller (IHMPC). The resulting nonlinear controller is shown to be stable when the IHMPC is globally stabilizing. The main advantage of the proposed controller in comparison to the IHMPC is a better performance as the model nonlinearities are taken into account in the computation of the control law. The advantage of the proposed controller compared to the existing dual mode NMPC is that constraints are also considered in the linear controller that is supposed to control the system when the state enters the terminal set. The performance of the proposed controller is compared to the stable IHMPC through simulation of an industrial styrene polymerization reactor.