In this paper, a novel three-dimensional (3D) irregular shaped geometry-based stochastic model (IS-GBSM) is proposed for sixth-generation (6G) millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) vehicle-to-vehicle (V2V) channels. To investigate the impact of vehicular traffic density (VTD) on channel statistics, clusters are divided into static clusters and dynamic clusters, which are further distinguished into static/dynamic single/twin-clusters to capture the mixed-bouncing propagation. A new method, which integrates the visibility region and birth-death process methods, is developed to model space-time-frequency (S-T-F) non-stationarity of V2V channels with time-space (T-S) consistency. The continuously arbitrary vehicular movement trajectory (VMT) and soft cluster power handover are modeled to further ensure channel T-S consistency. From the proposed model, key channel statistics are derived. Simulation results show that S-T-F non-stationarity of channels with T-S consistency is modeled and the impacts of VTD and VMT on channel statistics are analyzed. The generality of the proposed model is validated by comparing simulation results and measurement/ray-tracing (RT)-based results.