Increasing diagnosis capabilities in modern engine electronic control units (ECUs), especially in the exhaust path, in terms of emission and engine aftertreatment control utilize on-board NOx prediction models. Nowadays it is an established approach at hardware-in-theloop (HIL) test benches to replicate the engine's steady-state NOx emissions on the basis of stationary engine data. However, this method might be unsuitable for internal ECU plausibility checks and ECU test conditions based on dynamic engine operations.Examples of proven methods for modeling the engine behavior in HIL system applications are so-called mean value engine models (MVEMs) and crank-angle-synchronous (in-cylinder) models. Of these two, only the in-cylinder model replicates the engine’s inner combustion process at each time step and can therefore be used for chemical-based emission simulation, because the formation of the relevant gas species is caused by the inner combustion states. Incylinder models in HIL systems must respect the real-time criterion in addition to the computational effort of the electrical input/output signals of the HIL system. In order to meet these timing conditions, a zero-dimensional single-zone approach is currently widely being used and acknowledged.This paper will introduce a real-time-capable two-zone NOx model approach. The model is capable of replicating the dynamic behavior of the emissions, preparing for modern ECU algorithms and dynamic emission test procedures. Furthermore, it can be used as an extension to a common single-zone in-cylinder approach, yielding the simulation accuracy and robustness typical for HIL applications together with the capability for simulating emission output. By focusing on the hard real-time conditions together with the required parameterization and validation procedures, the paper will emphasize the applicability in closed-loop HIL projects.