What's the best way to answer "how good is this model" in the absence of any base model?

I have a classification model with imbalanced data so assuming all predictions are of the majority class doesn't seem much help as a baseline model.

class classification classification model data datascience good predictions