The problem with quant models, at least from whatever information I can pick up, seems to be that quants just assume that theirs is the only model out there and nobody else is doing what they do, when in reality, you have a bunch of other models, hence the joke:
"I just had a ten-sigma event happen to me! I haven't seen that in 3 months!"
As for which banks are making money, um, the good ones.