Fact is: there are lots of articles on arXiv and only about a third published traditionally (according to their statistics). Contrary to biology and medical science, where researchers are way more advanced in new publishing models (like PLoS and PeerJ, the second being almost green in flavour), in math and physics we don’t have any other option than arXiv, which is great, the greatest in fact, the oldest, but … but only if it had a functional peer-review system attached. Then it would be perfect!
It is hard though to come with a model of peer-review for the arXiv. Or for any other green OA publication system, I take the arXiv as example only because I am most fond of. It is hard because there has to be a way to motivate the researchers to do the peer-reviews. For free. This is the main type of psychological argument against having green OA with peer-review. It is a true argument, even if peer-review is made for free in the traditional publishing model. The difference is that the traditional publishing model is working since the 1960′s and it is now ingrained in the people minds, while any new model of peer-review, for the arXiv or any other green OA publication system, has first to win a significant portion of researchers.