This paper introduces three key features in geometry-based stochastic channel models in order to include massive MIMO channels. Those key features consists of multi-user (MU) consistency, non-stationarities across the base station array and inclusion of spherical wave modelling. To ensure MU consistency, we introduce the concept of "user aura", which is a circle around the user with radius defined according to the stationarity interval. The overlap between auras determines the share of common clusters among users. To model non-stationarities across a massive array, sub-arrays are defined for which clusters are independently generated. At last, we describe a procedure to incorporate spherical wave modelling, where a cluster focal point is defined to account for distance between user and cluster.