The contributions of long-lived nitrous oxide (N<sub>2</sub>O) to the global climate and environment have received increasing attention. Especially, atmospheric nitrogen (N) deposition has substantially increased in recent decades due to extensive use of fossil fuels in industry, which strongly stimulates the N<sub>2</sub>O emissions of the terrestrial ecosystem. Several models have been developed to simulate N<sub>2</sub>O emission, but there are still large differences in their N<sub>2</sub>O emission simulations and responses to atmospheric deposition over global or regional scales. Using observations from N addition experiments in a subtropical forest, this study compared six widely-used N<sub>2</sub>O models (i.e. DayCENT, DLEM, DNDC, DyN, NOE, and NGAS) to investigate their performances for reproducing N<sub>2</sub>O emission, and especially the impacts of two types of N additions (i.e. ammonium and nitrate: NH<sub>4</sub><sup>+</sup> and NO<sub>3</sub><sup>−</sup>, respectively) and two levels (low and high) on N<sub>2</sub>O emission. In general, the six models reproduced the seasonal variations of N<sub>2</sub>O emission, but failed to reproduce relatively larger N<sub>2</sub>O emissions due to NH<sub>4</sub><sup>+</sup> compared to NO<sub>3</sub><sup>−</sup> additions. Few models indicated larger N<sub>2</sub>O emission under high N addition levels for both NH<sub>4</sub><sup>+</sup> and NO<sub>3</sub><sup>−</sup>. Moreover, there were substantial model differences for simulating the ratios of N<sub>2</sub>O emission from nitrification and denitrification processes due to disagreements in model structures and algorithms. This analysis highlights the need to improve representation of N<sub>2</sub>O production and diffusion, and the control of soil water-filled pore space on these processes in order to simulate the impacts of N deposition on N<sub>2</sub>O emission.