In this paper, an optimal Generalized Predictive Control (GPC) is developed for binary distillation column using first principle model and linearized models. The nonlinear multivariable binary distillation column process is simulated with first principle differential equations and its linearized 16th order and reduced 5th order models were obtained. The GPC is designed based on original linearized model, reduced model and nonlinear first principle model. The performance of GPC with the above three considered model structures were compared. The response of GPC with reduced 5th order model shows similar characteristics of linearized 16th order model. Hence, the computation complexity can be reduced using a reduced order model for a binary distillation column process, without compromising on the performance.