In finite set model predictive control (FS-MPC) the computational burden increases due to the placement of each of voltage vectors in the cost function. In addition, the higher the number of levels results in the higher the computation rate. For this reason, by reducing candidates number for the cost function, the computational burden can be reduced. On the other hand, in the structure of the neutral point clamped (NPC) inverter, it is necessary to maintain the voltage balance between the capacitors. Balancing methods also increase the computational burden by adding an extra term to the cost function. This paper states a method to reduce the computational burden of a three-level NPC inverter in which the voltage balance of the capacitors is also considered. To reduce the computational burden in this method, only three voltage vectors are included in the cost function. Redundancy states have also been used to balance the voltage of the capacitors. The proposed method is evaluated using simulation.