If the features of Model $1$ are a strict subset of those in Model $2,$ which model will USUALLY have lowest TRAINING error?

Group of answer choices

It's impossible to tell with only this information

Model $2$

Model $1$