This paper deals with the reliability assessment of the flange cleat beam-to-column steel joints. Based on the set of tests of flange cleat beam-to-column joints the experimental bending resistance and stiffness has been obtained. Theoretical bending resistance and stiffness have been evaluated using existing and proposed (modified) mechanical models that were developed to account for the effect of stiffened cleats. Models uncertainties have been evaluated based on the comparison of the experimental and theoretical values. The limit state equations have been defined for each joint type and each theoretical model considered. The reliability indices for resistance and stiffness of joints were evaluated using FORM (First Order Reliability Method), SORM (Second Order Reliability Method) and AS (Adaptive Sampling). The obtained values of reliability index were discussed as well as the sensitivity factors of basic variables and sensitivities of reliability index to mean value and standard deviation of basic variables. Finally, a conclusion was based on the comparison between the results obtained from deterministic (semi-probabilistic) and probabilistic approach. The probabilistic approach demonstrated differences in respect to deterministic approach and pointed out the importance of using stochastic assessment for determination of the optimal theoretical model. From the probabilistic approach, optimal resistance models are EC3 model - method 2 for joint types with unstiffened top cleat and Faella model for joints with stiffened top cleat. Considering stiffness, Faella model is optimal for joint types with unstiffened top cleat, while for joints with stiffened top cleat no model is reliable enough. Nevertheless, because of its low reliability, EC3 stiffness model is not recommended for use with considered joint types. It is clear that stiffness of this type of joints requires further research and that the shortcoming of Eurocode in this issue should be solved in the second edition of the code.