K fold cross validation method-which model is exactly usful

Dear Akash sir and Jovian team,
In k-fold cross validation helper function,We returned 5 (5 because we splitted into 5 n_splits) different models to models list.
So exactly which model out of this 5 is good for testing and reporting purpose…last one or what?

Though we are splitting the batch into 5 splits, that means we are doing 5 splits with 5 sets of data. (5 x 5)

One set is taken as validation set for every splits. Although model based one split can perform better than others but for minimizing the bias (the model may carry) we take average of all the 5 splits.