One problem of many open-source models is that they are badly documented (there was a big discussion last year that many models used in scientific papers couldn't be replicated).

So reverse-engineering them is often harder than building your own from scratch.