This paper exploits idiosyncratic variations in school cohorts’ gender composition to investigate the short and long-term effects of school peers’ gender. Using French administrative data over the 2008–2012 period, it shows that the proportion of female peers’ in middle school not only affects students’ contemporaneous performance but also influences their subsequent educational attainment. More specifically, a larger share of girls among school peers increases girls’ test scores, reduces their dropout rates and increases their probability to graduate from high school several years later, especially in the scientific track. By contrast, it increases boys’ probability to attend a vocational school and decreases their high school graduation rate. I find suggestive evidence that these effects partially operate through a negative effect of opposite-gender peers on students’ school behaviour and through an adjustment of teacher behaviour based on the gender composition of the classroom.