Chloride-induced depassivation of steel reinforcement is one of the key durability issues for concrete structures. Conventionally prediction of service life using a probabilistic approach for chloride-induced depassivation relies on semi-empirical engineering models like the fib chloride model. However, such models use parameters such as the aging coefficient which are difficult to define in practice and introduce additional uncertainties that makes such models difficult to apply for realistic service life prediction. This study offers a new approach, employing a physics-based reactive transport numerical model in a probabilistic manner. The reactive transport model considers chemical and physical interactions within the concrete structure. Based on experimental data, a new binding isotherm for Portland cement paste is presented and used in the reactive transport numerical model. To demonstrate this approach, a case study compares the predictions made by the reactive transport model with those from the conventional fib chloride model. The reactive transport model predicts the chloride ingress after 6 years with less uncertainty compared to the fib chloride model. It also predicts the highest probability of depassivation after 50 years. Additionally, this study explores the impact of introduced uncertainties, given their substantial effect on the prediction results.