In this Malaysian case study, a full Bayesian approach is adopted to predict the posterior probabilities of carbonate lithofacies and porosity given the elastic parameters and density, which are inverted for from partial stack seismic data. We assume those parameters to follow a normal distribution for each facies and apply a Gaussian mixture model to compute conditional means and covariances. In a first step, the algorithm is applied on log data to predict facies from elastic parameters that can be compared with the real lithofacies from core. To estimate facies from seismic, an extracted trace at the wellsite, and an extracted inline have been used as input in 1-D, and 2-D, respectively. Results show that the algorithm can discern facies, which is determined by lithology and pore type successfully. The detection of thin beds and dolomite layers is precluded by low resolution and a smoothing effect of the seismic data, which does not capture the end members in the elastic range of facies. It can be inferred from the probabilistic result that the platform shows lateral facies changes, a fault-related zonation with respect to dolomitization, and complex margins from sequential co-sedimentation with siliciclastic strata.