In this paper we study the effect on the math gender gap from attending a coeducational school with single-sex classrooms versus attending a school with coeducational classrooms. That is, we compare the performance of girls versus boys within schools with single-sex classrooms compared to the performance of girls versus boys within schools with coeducational classrooms, using a difference-in-difference approach with school fixed effects. In line with the results of the previous literature that has found positive effects for female students, we find that coeducational schools with single-sex classrooms reduce the math gender gap by more than half. The effect is consistent with an increase in the math achievement of female students with no decrease in the achievement of male students. Finally, this effect is not driven by teacher characteristics or student-teacher gender matching, but it seems to be driven by how female students respond to the gender composition of the classroom.