Precision

The proportion of positive predictions that are actually correct. High precision means that when the model says something is positive, it is usually right, but it may still be missing many true positives. Precision is most important in applications where false alarms are costly, such as content moderation or medical diagnosis.
See also: recall, f1 score.