Jean-Patrick Lebacque IFSTTAR GSOM models combine a kinematic macroscopic description of traffic with driver specific attribute dynamics. Driver attributes can be passive (OD, path, vehicle and/or driver type) or active (vehicle state, information, stochastic perturbations). It will be shown that these models can be adapted to describe complex flows such as associated with complex geometries of the infrastructure (multilane, intersections). For instance it is possible to combine GSOM models with lane assignment models. Other aspects to be discussed are large network modelling and numerical aspects, notably those related to the lagrangian Hamilton-Jacobi formulation of the model.