As an important form of classroom interaction, peer interaction plays a very important role in group learning. It can promote students’ development of high-level thinking ability and construction of knowledge. But there are two viewpoints for constructing group member. The first is that heterogeneous group is benefit to the development of students and the second is that homogeneous group is benefit to the development of students. The purpose of this study is to find out if the group member construction of peer interaction influences creative problem finding ability (CPFA) and if this influence is related to students’ ability level and task difficulty. This study utilized a 2 (teaching materials or tasks of different difficulty: high and low) × 4 (group member construction: homogeneous group, heterogeneous group, voluntary group and control group) mixed design. Before the experiment, 148 primary school students of Grade 4 took part in the CPFA Test and then according to the CPFA score, they were matched into 4 equal groups and studied two materials with different difficulty and creative problems were proposed after the study. The results showed that: Compared with the independent learning, peer interaction learning could promote the development of CPFA more effectively; Group member construction had significant influences on the students’ CPFA; Material (or task) difficulty had significant influence on the students’ CPFA; there was significant interaction between group member construction and material (or task) difficulty. For the lower difficult material, homogeneous group was the best, voluntary group was better than control group and heterogeneous group was the worst. For the higher difficult material, voluntary group was the best, homogeneous group was better than heterogeneous group and control group was the worst. Lower-ability students had best performance in heterogeneous group; middle-ability students had best performance in homogeneous group; higher-ability students had best performance in voluntary group.