The paper is focused on the calibration of non-linear 3D finite element (FE) analyses to simulate punching failure of R/C flat slabs. The calibration procedure is developed with reference to the code ABAQUS, which is one of the most used computer codes in nonlinear modelling of R/C structures. Generally, the calibration of a nonlinear FE model is grounded on one test only, so its reliability could be limited. Here a hybrid method for the calibration of FE models of R/C flat slabs failing in punching is proposed and discussed. The method consists in calibrating input data by comparison of finite element model (FEM) results with both experimental data and predictions provided by analytical models. The procedure allows for a consistent calibration to be performed, valid for a wide range of longitudinal reinforcement ratios, from 0.5% to 2.00%, and concrete grades, from C20/25 to C50/60. A case study is investigated using the proposed method. Results show that calibrated values of the fracture energy lie between those provided by Model Code 1990 and Model Code 2010. From the new calibration procedure, a relationship between fracture energy and concrete compressive strength is also derived and blind analyses are performed to check its reliability against experimental results.