The heat transfer in liquid metal–cooled rod bundles is modeled with a knowledge-based best-estimate system code. Thereby, the focus is on the heat transfer enhancement due to flow perturbations. These perturbations are caused by local geometrical variations, such as sudden expansions and contractions, in the flow channel. The accurate calculation of the heat transfer is important for the safety demonstration of, e.g., subassemblies. Safety-related parameters, such as fluid and wall temperature, have to satisfy certain limits during normal and off-normal operation as well as during accidents. Up to now, fully developed flow is assumed for heat transfer in liquid metal–cooled rod bundles. The effects of local heat transfer enhancements were ignored in best-estimate system codes. The currently used empirical heat transfer models are functions of the Péclet number only. Several experimental and numerical investigations show that flow perturbations induce higher heat transfer due to increased turbulences, accelerated flows, and secondary motions. In this paper, the effects of the entrance region and the presence of spacer grids on the heat transfer are investigated. Empirical models for that are selected and applied. These empirical models are functions of the Péclet number, the geometrical perturbation, and the distance from the perturbation in the flowing direction. The calculated heat transfer coefficients at the bundle entrance and in the vicinity of spacer grids are twice as high compared to bare rod bundles under a fully developed flow condition without any flow perturbation. Because of the higher heat transfer, lower wall temperatures are to be expected. This provides additional safety margins during normal and off-normal operation as well as during accidents. Furthermore, the considerable increase of heat transfer shows that existing perturbations have to be considered to obtain accurate and reliable results.