This paper assesses the effect of teachers on the gender gap in student test scores. It combines different empirical strategies from the value-added and labor economics literature to estimate teacher value-added and its contribution to the math and reading gender gaps. We use rich administrative data from Chile, that allows us to follow teachers through different classes in different years. Our main findings indicate that teachers explain up to 18% of student test score variance and help reduce the gender gap in math by 16.9%. The reduction in the math gender gap is greater in voucher schools (16.1%), among students with more educated mothers (24%) and among those with female math teachers (32.2%). We provide evidence supporting a within-class effect instead of sorting (between-class effect). We conduct several tests and robustness checks to assess the reliability of our findings.