In this work, a model of a semibatch poly(ethylene terephthalate) reactor in which dimethyl therephthalate (DMT) and ethylene glycol (EG) react during the transesterification stage is developed and used for an application of nonlinear model predictive control (NMPC). The reactor is connected to a packed-bed column with a partial condenser. In the model, the hydrodynamic behavior of a column section is described by a correlation valid for hydrodynamic regimes under the loading point. The description of the heating of the reactor includes the division of the equipment into different compartments, each with a given thermal inertia and with a thermal resistance assumed between them. The column is considered dry at startup. NMPC is applied to the system model, and its performance is compared to that of an optimally tuned PID controller on three different temperature trajectories. The performance of NMPC is better than that of PID. The NMPC performance is studied for different modeling gaps, and it is not seriously degraded for the cases studied. This study does not intend to prove NMPC superiority over PID in the control of a semibatch polycondensation reactor. The application of NMPC illustrates the main difficulties involved in the control of the process, as the model presented here is able to describe the main features of its dynamics.