Peer assessments have demonstrated predictive validity for many military outcomes, including academic standing and leadership, success in training, and combat effectiveness. Though military training is often conducted at the squad or platoon level, little research has examined the effects of group size on the reliability and predictive validity of peer assessments for performance outcomes. The present study used archival data to examine the impact of group size on the interrater reliability and the predictive validity of peer rankings for leadership performance. Average peer rankings at the squad and platoon level were examined as predictors of performance outcomes for 191 junior Army leaders. Results indicated that platoon-level peer rankings tended to be better predictors of instructor-rated leadership scores in a garrison environment, though findings were mixed when examining leadership in field training exercises.