Discriminating similar molecules remains a very challenging problem for semiconductor gas sensors. Here, we report a method to achieve precise gas discrimination of similar chemical vapors (mesitylene, o-xylene, and toluene) by using cross-reactive arrays consisting of metal oxide semiconductor and graphene sensors. It is difficult to identify these three chemicals as they have very similar responses to these sensors. Through a cross-reactive Principal Component Analysis (PCA) of the sensor response features, however, the discrimination accuracy improved from about 70% with a single gas sensor to almost 100% with the cross-reactive sensor array. Such a precise discrimination and the low-cost planar process make this approach a very attractive candidate for smart gas sensing and for future Internet of Things (IoT) applications.