In this paper, we analyze flow level performance of mobile users in 4G/pre-5G cellular networks such as LTE and LTE-A. To this end, we develop a two-levels model that provides users' performance in a cell and end-to-end performance in the network. At a cell-level, the model is a multi-class PS queue that captures mobility of users between zones of a cell, through a simple mobility model, that is decoupled from the cell model itself, enabling to directly apply the approach to more realistic mobility patterns. At a network-level, the model is a simple Discrete Time Markov Chain that reproduces the routing of mobile users between the different cells of the system. We first show that this model is consistent with known analytical bounds corresponding to a system with either static users or users having an infinite speed. The outcomes of our model confirm that mobility may improve both users and cells performance, and enable to quantify the gain. The model also shows that, while inter-cell mobility balances the load between cells, it does not lead to such improvement of throughput as intra-cell mobility.