Measuring the gender peer effects on student achievement has recently attracted a lot of attention in the literature. Yet, the results are inconclusive. A substantial amount of research shows that having relatively more girls in a division increases the academic achievement of all students. Nevertheless, the identification of pure gender effects remains a challenge due to the fact that girls outperform boys in overall academic performance. Our study overcomes this identification problem in a setting where girls are not academically better. Using 2009-2010 school year data on 8th graders in Turkey, this paper disentangles pure "academic" peer effects and "gender" peer effects. Our estimations reveal that the higher the share of females in a division, the lower the likelihood that a student drops out. One standard deviation increase in the share of females in the division decreases the likelihood of dropout by 0.3 percentage points. This result holds even though females are 9.32 percentage points more likely to drop out. These findings are robust to the inclusion of various control variables e.g. parental and academic background of the student, school and regional characteristics. We also find that the gender peer effects are prevalent in both females and males.