All models are wrong. There is no such thing as a perfect model outside of trivial classroom models e.g. spherical cow. Modeling fluid dynamics for aerodynamic lift, structural integrity models for bridges and buildings, etc . all have errors. They don't account for all variables and it is impossible to do so.
Science isn't built on models. Models are built on science. As with any other branch of science models are used to help get a better understanding of the phenomena being studied. Models are TOOLS that are built out of the results of science.
As to GCM's in particular, there is plenty of information out there describing the models, their error bounds, what they account for, what they don't, so on and so forth. For a layman's summary the IPCC does a fairly decent job describing the models, what they're used for, and accuracy.