Knowledge co-creation between enterprise-dominated online innovation communities and external individuals is a pivotal path for enterprises to improve innovation performance. However, little is known from existing research about how initiator characteristics, content characteristics, and community dimension affect co-creation performance in different co-creation settings. Therefore, based on 150 co-creation proposal samples from the Xiaomi online community, we use configurational and contingency theory to construct a theoretical framework and reveal configuration paths for shaping desired co-creation performance. The findings indicate that the multidimensional influences of knowledge co-creation can be combined into nine configurations to yield high or low co-creation performance in different co-creation settings. Across all solutions, ability and community responsiveness are the most common core conditions, and their complex interactions with other elements jointly affecting co-creation performance. Furthermore, these configurations are context-dependent, namely, the co-creation settings exert a moderating role between the multidimensional influences of knowledge co-creation and co-creation performance.