The role of criminal, social interactions occupies a central place in criminology, yet minimal research exists on the relationship between co-offender networks and dimensions of offending. Drawing on the social network literature, this investigation hypothesizes that a link exists between the level of redundancy (i.e., the extent of overlap) in an individual's co-offender network and offending versatility. Relying on longitudinal data for a random sample of delinquents from Philadelphia, this study begins by constructing egocentric co-offending networks for the respondents. Then, using Tobit regression models, it finds that higher levels of co-offender network redundancy (more dense networks) are related to higher levels of specialized offending in group crimes, but no such relationship exists with overall (i.e., solo and group) offending specialization. The discussion considers the implications of these findings and offers suggestions for future research.