Children begin to continually improve their self-evaluation ability through social comparison during middle childhood. Previous studies indicated that there is no agreement on whether perception of peer acceptance differs when children are rating a same-gender or opposite-gender peer group. Furthermore, some studies indicated that children develop more accurate and positive bias perceptions of themselves with increase grade in middle childhood. This issue is important in part because there might be an association between children’s perception of peer acceptance and their social withdraw behaviors. Therefore, the main goals of the present study were to investigate gender and grade differences in peer acceptance perception accuracy in middle childhood, and to explore associations between these perceptions and two types of social withdrawal (active isolation and passive withdrawal). The study sample was selected by a random cluster sampling method, comprised of students from 3rd grade to 6th grade in an elementary school in Wuhan City, China. A total of 498 students (276 boys and 222 girls) completed the questionnaires. The children in the current study were asked to rate how much they liked their classmates and how much they would be liked by each classmate, and children also were asked to complete the Self-Perceived Social Competence (SPSC) measure and the Revised Class Play questionnaire. Accuracy of peer acceptance was indexed by the mean difference (absolute values) between pairs of predicted and received ratings, and bias of peer acceptance was measured by standardized residuals derived by regressing children’s responses to the Social Acceptance Subscale of the SPSC onto peer acceptance ratings. F-test was performed to examine sex or grade differences on the variables, and multilevel regression analyses were conducted for explored the association among variables. The results showed that: (1) Higher grade children displayed more accurate peer acceptance perception (including perceptions towards both genders) of peer acceptance than that of lower grade children. (2) There was no gender difference in the accuracy of perceptions of peer acceptance by same-gender groups, but girls in fourth grade, compared to boys in fourth grade, displayed more accurate perceptions of peer acceptance by the opposite-gender group. (3) Same-gender and opposite-gender accuracy of peer acceptance significantly positively predicted both active isolation and passive withdrawal behaviors; peer acceptance perception bias significantly negatively predicted passive withdrawal, but not active isolation behaviors. In terms of the study’s unique contribution to the literature on the first time to used Chinese sample to investigate accuracy and bias of children's perceptions of peer acceptance, and the first time to explore the association between two types of social withdrawal and them.