Predicting the singlet-triplet splittings of divalent radicals is a challenging task for electronic structure theory. In the present work, we investigate the performance of multiconfiguration pair-density functional theory (MC-PDFT) for computing the singlet-triplet splitting for small main-group divalent radicals for which accurate experimental data are available. In order to define theoretical model chemistries that can be assessed consistently, we define three correlated participating orbitals (CPO) schemes (nominal, moderate, and extended, abbreviated as nom, mod, and ext) to define the constitution of complete active spaces, and we test them systematically. Broken-symmetry Kohn-Sham DFT calculations have also been carried out for comparison. We found that the extended CPO-PDFT scheme with translated on-top pair-density functionals have smaller mean unsigned errors than weighted-average broken-symmetry Kohn-Sham DFT with the corresponding exchange-correlation functional. The accuracy of the translated Perdew-Burke-Ernzerhof (tPBE) on-top pair-density functionals with ext-CPO active space is even better than some of the more accurately parametrized exchange-correlation density functionals that we tested; this is very encouraging for MC-PDFT theory.