Rock properties are useful for safe operation and design of both surface and underground mines including civil engineering projects. However, the cost and time required to perform detailed assessments of rock properties are high. In addition, rock properties are required in numerical modeling. Different models have been proposed for quick and easy assessments of rock properties but majority of these models are regression-based, which are incapable of capturing inherent variabilities in rock properties. Therefore, this study proposed three different soft computing models (i.e., double input-single output ANN, multivariate adaptive regression spline, genetic algorithm) for accurate prediction of several mechanical properties of coal and coal-like rocks. The performances of the proposed models were statistically evaluated using various indices and they were found to predict rock properties suitably with very strong statistical indices. The proposed models were validated further using external datasets aside from those used in the model development to test the generalization potential of the models. The Pearson's correlation coefficients for the validation were close to 1, indicating that the proposed models can be used to assess geo-mechanical properties of coal, shale, and coal-bearing rocks.