High-quality data sets of free carbenes have remained unavailable in the scientific literature so far. We provide approximately 5k and 8k verified carbene structures in their respective singlet or triplet state. Vertical spin gaps have been computed at higher order multireference level of theory (MRCISD+Q-F12/cc-pVDZ-F12). The carbenes presented are all derived through double hydrogen abstraction from saturated carbon centers of a subset obtained by randomly sampling the chemical space of approximately 300k carbene candidates possible within closed shell organic molecules in the QM9 data set.