The only "model of a whole brain" we have currently is the one of C. elegans, and even that's debatable because it does not model chemical processes which modulate neural activities, and therefore does not provide an accurate input/output mapping (in general).
No, it doesn't predict activity from parameters.
> C. elegans, and even that's debatable because it does not model chemical processes
"the map is not the territory". This sort of criticism isn't even scientific
Not sure what you mean.
For example, we have a whole brain model of rest state dynamics that can represent aspects of data (eg graph metrics such as degree) from multiple sclerosis patients, distinctly from healthy controls. Here, the map is the model parameters, and the territory is the patient's actual brain. The map is deemed useful if its structure represents that of the territory.
http://www.scientificamerican.com/article/the-root-of-though...
My day job is building and evaluating such models; we have one on seizure propagation that is entering clinical trial.