This article presents an applicable real-time thermal model for the temperature prediction of permanent magnet synchronous motors. The load capacities of most permanent magnet synchronous motors are usually limited by the temperature, and overheating is one of the main reasons for permanent magnet synchronous motors breakdown, so an applicable temperature prediction approach is helpful to improve motor utilization and protect permanent magnet synchronous motors from thermal distortion. Compared with embedding temperature sensors into motor structures, implementing real-time thermal model in motor controllers is a cost-effective and rapid response protection method, but it still faces the challenges on the temperature estimation accuracy, the complexity of the model parameters and the computational efforts. To balance every aspect of these challenges, this article tries a simple real-time thermal model to accurately predict the thermal behavior by elaborately modeling stator core losses and considering motor itself cooling ability. The affections of the motor current and speed on the core losses are analyzed and a polynomial equation is adopted to deal with their dependencies. To simulate the motor speed impact on the cooling ability, motor speed is involved in the variable thermal conductance of the motor housing inside the surroundings by another polynomial equation. This article describes how to get the most parameters of the proposed real-time thermal model through motor basic dimensional information and introduces the test methods employed to determine the parameters of the above two polynomial equations. In the experiments, first the thermal model building process is provided by an actual permanent magnet synchronous motor with two simple tests, and then the online analytical expressions with the obtained parameters are implemented in the drive controller to verify the performance of the proposed real-time thermal model. The results of the performance tests show that the real-time thermal model has a good agreement between estimated and measured temperature values, and its performance can satisfy the most actual applications.