This paper tackles the challenging task of developing accurate easy-to-use analytical models for performance evaluation of WiMAX networks. The need for precise fast-computing tools is of primary importance to face complex dimensioning issues of WiMAX cells. Here, we present generic Markovian models developed for the rtPS service class defined in the WiMAX standard. This service class is dedicated to real-time applications with constraints on both their bit rates and the latency of their packets (i.e., streaming). Our analytical modeling is performed in two steps. First, we focus on the connection level only considering the bit rate constraint. Then, we extend the model to the packet level to formulate the packet delay QoS parameter. The resulting models can account for multiple traffic profiles at both levels while always keeping an instantaneous resolution. The models are compared with simulations that show their accuracy. Finally, we show how our models can be used to solve dimensioning issues while avoiding days of simulations.