It was unclear what level of reliability was provided by appliance of the Eurocode load models for fatigue due to train loads, meaning that it might be insufficient. Therefore, it was investigated what the provided reliability is, compared to what is should be. Within this report, the focus is solely on fatigue induced by bending moments on reinforced concrete sections. Determination of reliabilities is done by creating a fictitious design, exactly satisfying the Eurocode. Then, this design is used as input for a probabilistic analysis, where the load models are replaced by measured traffic, and other uncertainties were quantified. The target reliability for the fatigue limit state proved to be vague in the Eurocode. However, the target reliability for ultimate limit states can be adopted as an upper bound, which was used in this report. The reliability was shown to be below this target, caused by, among others, a partial factor which is equal to 1 for fatigue loads according to the Eurocode. The idea was that variations in loads will converge due to the vast number of appliances. However, here it is argued that correlations between variations, which systematically affect each load its effect, invalidate this assumption. Also, reliabilities were shown to depend heavily on the span lengths (remarkably lower for spans of 1-10 m. It was demonstrated that such behavior results by lack of variety in the load model axle distances and loads, compared to measured traffic. Recommendations are, apart from further research, to increase the partial factor for fatigue loading. Also, a novel load model should be created with comparable variation in axle distances and loads as measured traffic.