This paper compares the spatial cross-correlation functions (CCFs) and normalized channel capacities of four recently proposed massive multiple-input multiple-output (MIMO) channel models. Three of them are geometry-based stochastic models (GBSMs) and one is Kronecker-based stochastic model with birth death process on the array axis (KBSM-BD-AA). In addition, the impact of the elevation angles in the three dimensional (3-D) twin-cluster model and the 3-D unified GBSM on the resulting channel capacities, as well as the impact of polarization antennas in the 3-D unified GBSM on channel capacities, is investigated. Simulation results show that elevation angles can cause large impact on channel capacities and polarization antennas can halve the dimension of an antenna array at the cost of channel capacity loss.