Lately this has been a hotly contested issue, with arguments both for and against open sourcing code. Though it is desired that code and weather models be open sourced. The incentive mechanisms that researchers in this field work under act against this goal. It is not uncommon to guard model, data and code, at least temporarily till enough publications have been wring out of them. Moreover the community is not very tolerant of mistakes, which makes people shy away from open sourcing their code lest embarrassing mistakes are discovered. All in all not a very healthy atmosphere.
Nonetheless many are working towards the goal of open sourcing understandable models (in python) http://initforthegold.blogspot.com/2008/08/climate-models-is... But I do not know what has come out of it.
Other places of interest are http://mitgcm.org/ and http://wrf-model.org/users/release.php were you can actually checkout code.
Disclaimer: I have not played with any of these code. Its something that interests me, yet something that I have to put aside at the moment.