Previous studies have confirmed the benefits of small group sizes in asynchronous online discussions. However, there is no consistency in which small group size is the optimal option for maximizing students' discussion performance. Moreover, few researchers attempted to apply the smallest group size (i.e., one-on-one) to online discussions and examined its effects. This study focuses on comparing two small group size choices regarding students' discussion performances. 21 graduate students formed pairs in the first half of the semester, and in the second half, they formed groups of three to five. Paired-sample t tests were conducted to detect the performance differences between the two group sizes. The results showed that students significantly performed better in contributing interactions, providing information, engaging in social conversations, demonstrating cognitive engagement, and constructing knowledge. This study provided a new perspective of designing discussion activities in online forums.