We explored how gender and gender similarity affects friendship dissolution following the transition to middle school. We predicted that both gender and gender similarity (measured by perceived similarity to own-gender and other-gender peers) explain dissolution trends and that less own-gender similarity or more other-gender similarity predicts more friendship dissolution. We considered gender and gender similarity at both the individual and dyad level (reflecting the discrepancy between friends). Participants were 198 students in Grade 6 (42% Latinx, 21% Caucasian, 10% Native American, 8% African American, 2% Asian American, and 17% mixed backgrounds; 77% qualified for free/reduced meals) in reciprocated same- or mixed-gender friendships followed from fall to spring of the academic year. Multilevel multimember logistic regression models, nesting friendships within each participating individual, demonstrated that girl-girl friendships were less likely to dissolve than boy-boy friendships, and mixed-gender friendships did not dissolve more than same-gender friendships. Feeling similar to one's own gender predicted less dissolution, but feeling similar to the "other" gender did not increase friendship dissolution. There was no support for the hypothesis that feeling similar to both genders (i.e., androgyny) protected against friendship dissolution, nor was there any support for the hypothesis that dyad-level differences in gender similarity would predict dissolution. The discussion focuses on the importance of conducting individual and dyad-level analyses as well as including gender similarity constructs when studying gender differences in friendships and their trajectories over time.