Depending on the magnitude of the model - I've used Access to provide the model simulation. If it's large in magnitude and records, I personally use MatLab (w/SimulLink) to great extent. I've used Stella in the past with varied success. The reason is the types of PDFs and CDFs that you might want to use are easier to work with since they are built in to those packages. I also like to run a large data set to see the output pattern to provide a feedback mechanism for variable clarification - this can be tasking for applications that aren't built for this purpose.
In all, it just depends on how exacting the model needs to be and the tolerance to compute the confidence level.
Whichever path you choose - the advice I have is to create and generate your random tables to feed your model in advance so the simulations do not take much time (you will only be referencing a data set at sim time, not generating and referencing).
-dK