Abstract: Traditional models for shear strength of reinforced concrete (RC) beams are generally deterministic models and exhibit low computational accuracy and large numerical fluctuation, due to the fact that they do not take into account the aleatory (physical) uncertainties of various parameters such as geometry, material properties and boundary conditions as well as epistemic (model) uncertainties of the modelling. Based on the modified compression field theory (MCFT) and the critical crack angle model considering the influence of shear span ratio, a deterministic model for shear strength of RC beams was established first. Subsequently, a probabilistic model for shear strength of RC beam was developed by using the Bayesian theory and the Markov Chain Monte Carlo (MCMC) to take into account the influences of both epistemic and aleatory uncertainties. Finally, the applicability and efficiency of the proposed probabilistic model were validated by comparing with experimental data and traditional deterministic models. Analysis results show that the proposed probabilistic model is of good accuracy and adaptability. The model not only can describe the probabilistic distribution characteristics of shear strength of RC beams, but also provide a benchmark to calibrate the confidence level of traditional deterministic models and provide an efficient way to determine the characteristic values of shear strength of RC beams with different confidence levels.