I'm not an expert in this field but I had the same question about a year ago...
First, I think that there is a piece of information you need to give us when you ask this: what approach do you use? RANS? LES? Because, for exemple, those two approach are quiet different and are completly different in term of turbulence modeling.
Second, reading on the topic, I became aware that it really depends on your application and conditions of simulations. So the litterature and formulation of models can help you avoid some of them which, perhaps, will not be pertinent in your case.
After all of this, if there is more than one left, I would say that there is no other choise than testing them and conclude by yourself.
But, I think you should give us more details as I said to help people give you an orientation.
A stochastic lagrangian model with a lattice-boltzman scheme achieves very good results. There is a good article about this titled “Modeling of turbulent gas–liquid bubbly flows using stochastic Lagrangian model and lattice-Boltzmann scheme” by Sungkorn et al.