[sharperguy]

At this point, with any new model I think it makes sense to wait until you can run the model on your own input before making any assumptions based on cherry picked examples.