Why do inequalities in schooling persist, even in relatively egalitarian school systems? This article examines within school sorting as an explanation. We use classroom data on friendship networks in 480 European secondary schools and contrast comprehensive (England, Sweden) and tracked systems (Germany, Netherlands). Our question is to what extent comprehensive systems reduce achievement sorting at the level of (a) schools, (b) classrooms, and (c) friendships. Between-school variance in achievement is lower in comprehensive systems. However, this is counterbalanced by greater sorting within schools, between classrooms, and, especially, in friendship networks. Still, comprehensive schools create more equal environments for two reasons. First, the difference in between school sorting is larger than the difference in within school sorting. Second, within school sorting is less strongly related to social background characteristics. These findings help explain both why comprehensive schools produce more equal outcomes and how substantial inequality can nevertheless persist.