Electric vehicle sales have accelerated in recent years due to a wider customer acceptance. However, the still limited driving range continues to be a barrier to purchase for many customers. To improve the driving range, it is particularly important to further reduce all losses in the powertrain. This applies in particular to the temperature-dependent motor and inverter losses. In this context, this paper presents a high-fidelity motor model based on MotorCAD and ANSYS Maxwell, which is thermally controlled using an economic Model Predictive Control (MPC) approach to reduce the temperature dependent losses. A detailed explanation of the high-fidelity motor model is given, followed by a system-level validation including the thermal system model. The setup is used to simulate three different cycles, namely a highway drive, a rural road drive, and a long urban drive. A comparison between the MPC, which actively controls the rotor, winding and the inverter junction temperature, and a rule-based strategy is used to analyse the motor-level losses in detail. While the total MPC savings system level are up to 2.86% at 35 °C ambient temperature, 0.82% of this is saved due to increased temperatures of the motor caused by reduced cooling while driving first highway and then into the city. For a pure highway cycle the motor savings increase up to 1.12%. One major outcome, which is enabled by the detailed modelling approach, is that the majority of those motor savings are originating from reduced ac-copper losses at elevated temperatures. The reason was found to be a lower magnet flux density and a higher winding resistance which both lead to less eddy current losses in the winding. Moreover, the inverter losses were reduced by cooling the inverter prior to acceleration from standstill. By using the NTC behaviour of the IGBTs in low current region, temperature-dependent savings of up to 0.56% were achieved.