In the talk I will first recall the basic ingredients of the class of second order fluid models of traffic flow introduced several years ago with Aw : the "AR" model, sometimes called the "ARZ" (for Zhang) model in the traffic flow community. In particular, I will stress some more recent comments on the Lagrangian features of the model and in which situations (if any !) it is worth (?) using such models, (slightly) more complex than the LWR models. I will then describe a recent work with M. Herty and S. Moutari on the implications for modeling junctions with a criterion of maximal flux. The overall result is that the "more agressive" drivers take most or even all of the available supply on the outgoing road(s), which of course (a) sounds pretty reaonable! and (b) has no reason to be true for a first order model.