In this article, we have assessed the performances of some recently proposed density functionals for the prediction of reaction energies involving radicals, notably bond dissociations of small organic molecules or of TEMPO-based ones, and beta-scissions, focusing on our TCA family and on range-separated hybrids. It is found that no functional belonging to these two families is able to compete with the M0x one. We have tried to improve the performances of the range-separated hybrids by the optimization of the attenuation parameter, but the improvements for one dataset lead to an unavoidable deterioration for the others. Furthermore, the differences between two different approaches to the long-range/short-range separation are discussed in terms of average enhancement factors, emphasizing the crucial choice of the approximate scheme used for the short-range part. Finally, the influence of the geometries has been considered and found to be negligible for this kind of molecular sets, validating the usual single point energies strategies developed in such benchmarking assessments.