Leader-member exchange (LMX) differentiation, the extent to which the leader forms different quality exchange relationships with group members, is the fundamental premise of LMX theory. The present study tests whether, why, and when LMX differentiation increases group outcomes. More specifically, I suggest a dual perspective model of LMX differentiation and group effectiveness, which includes both leaders' and members' perspectives on LMX differentiation. To explicate the effects of leader- and member-rated LMX differentiation on group effectiveness, I delineate two group processes (task and socio-emotional processes) to link LMX differentiation to two aspects of group effectiveness (group performance and group viability). I propose that leader-rated LMX differentiation increases group performance via task group processes (group role clarity and group coordination) whereas member-rated LMX differentiation decreases group viability via socio-emotional processes (relationship conflict and group potency). Furthermore, I propose moderators (distributed expertise, leader-leader exchange, leader prototypicality, and group power distance) suggesting when leader-rated LMX differentiation relates to task group processes more strongly and positively, and when member-rated LMX differentiation relates to socio-emotional group processes less strongly and negatively. Survey data, based on the US and South Korean employees working in 57 teams, were collected from three different sources (group leaders, members, and upper-level managers) at two points in time (with 3-month time lag). Results revealed three important findings: (a) leader-rated LMX differentiation is positively correlated with group role clarity, group coordination, and group potency and negatively correlated with relationship conflict, (b) group mean member LMX and group mean leader LMX were more robustly, positively related to most group process and effectiveness variables, and (c) leader prototypicality moderates the relationship between member-rated LMX differentiation and relationship conflict such that the negative relationship is significant only when members perceptions of leader prototypicality was high. The present study contributes to the LMX and team leadership literature by (a) providing a theoretical framework of a dual perspective to understand the LMX differentiation-group effectiveness relationship at the group level, (b) providing empirical evidence showing that leader-rated LMX differentiation is positively related to some group processes although these relationships were not significant when controlling for group mean LMX, (c) finding that the leaders' and members' perspectives on LMX differentiation are not highly correlated, and (d) emphasizing the role of leader prototypicality in understanding the effect of member-rated LMX differentiation at the group level. Based upon these findings, I suggest future research directions such as ways to improve measurement and operationalization of LMX differentiation, development of a theoretical model explaining the low level of LMX agreement between members and leaders, and examining LMX differentiation in broader contexts (e.g., HR systems or strategy and cultural contexts).