This paper presents the design of a model predictive controller for grid-tie multilevel photovoltaic inverters. The proposed control scheme has a cascade-free structure and it uses a simple cost function optimization algorithm to compute the control action to be applied at the next sampling period. The cost function is designed by merging the grid current tracking with the overall dc-link voltage control and the neutral-point voltage minimization, avoiding the extra linear controllers used by conventional schemes. The strategy is validated in a three-level inverter connected to the mains trough simulation results using measured irradiance and temperature profiles.