Nuclear waste management is a major challenge for the nuclear industry. Radiological characterization of nuclear waste packages is an important step of this industrial process as an inaccurate analysis could lead to the over classification of a package and, consequently, to a higher cost of storage. In this context, INNOWTECH and CEA List develop a low-cost and automated nuclear measurement method based on gamma spectrometry carried out with CdZnTe detectors in view of reducing uncertainties on radionuclide activities. In a previous study, we showed that the MCNP6 model of the GR1+ KROMEK CdZnTe detector created by Selivanova et al. (Appl. Radiat. Isot., 2019) contains non-modelled parts and inaccurate parameters that contribute to reducing the validity range of the simulation model. In this paper, we carry out an optimization study of such a Monte-Carlo simulation model using an original approach. First, we assess experimentally the detector’s response according to the angular distribution in metrological conditions at LNHB (Laboratoire National Henri Becquerel, France). Secondly, we conducted a comparative study of different gamma spectrometry softwares and retain the sIGAle code developed by CEA List. Thirdly, we fine-tuned the MCNP6 simulation model. On the one hand, the geometry of the model is adjusted thanks to gamma radiographies of the CdZnTe detector conducted with a pixelated CdTe detector (Widepix system) and an 241 Am source. On the other hand, a tool named SBINI (Simulation-Based Inference for Nuclear Instrumentation) is used to infere physical parameters of the MCNP6 model in order to optimize simulation results according to experimental ones. This new approach based on two complementary tools can be applied to optimize Monte-Carlo simulation models of various particle detectors.