XGBoost multi classification objective and eval_metric

Question

So from what I understand the multi:softmax and multi:softprob differ in what they return, namely softmax gives you back the best label while softprob gives you back the pro