After downloading, it looks like it uses a custom "talkie" model architecture that is not supported at least by llama.cpp.

Edit: This one has now been deleted as well for some reason...