In this study the processing routes of three different stainless steel grades were followed; during the reducing period in the AOD plant samplings of steel and slag and measurements of temperature and oxygen activities were performed. The goal was to develop tools for the improvement in the comprehension about the development of the non metallic inclusions in stainless steels. The inclusions in the steel samples were analysed by means of a SEM and then the evaluation of the oxide activities and phases present within them was performed with Thermocalc^©. A model developed by Janke and co-workers was implemented in order to compute the activity of elements in the liquid steel. This is based on Wagner's formalism but it is more accurate for the steel baths where solutes are not at infinite dilution. This is the reason why Janke's formalism is better than the Wagner's one in the study of stainless steels, as they are high-alloy steels. It was found that Janke's formalism lacks reliability, when it neglects some strong interactions between two elements (e.g. O and another element having high affinity to oxygen).The activities of oxides at the equilibrium with the liquid metal were calculated and compared to those obtained from the Thermocalc^©-based analysis of inclusions. From this comparison some hypotheses were made about the nucleation and modification of SiO₂, Al₂O₃ and Cr₂O₃. By making some variations to the model for the calculations of the oxides activity at the equilibrium with the steel bath, it has been also assessed how the model results are influenced by the choice of some thermodynamic parameters such as the interactions coefficients and the equilibrium constants.