Many believe that classroom interactions play an important role in students’ academic achievement, but there is little evidence on peer effects within subclassroom groups. We exploit random seat assignment in a Chinese middle school to estimate how the gender of neighboring students affects a student’s academic achievement. We find that being surrounded by five females rather than five males increases a female’s test scores by 0.2–0.3 standard deviations but has no significant effects on a male’s test scores. These results suggest a low-cost way to potentially improve performance within the world’s largest school system.