Vertical mixing parameterization is a major source of uncertainty in ocean circulation models. In recent years, deep learning has been increasingly applied to improve traditional parameterization schemes due to its ability to capture complex nonlinear processes. This study integrates the deep learning-based vertical mixing parameterization (DLVMP) proposed by Fang et al. (2025) into the Climate system ocean model (LICOM) developed by the Institute of Atmospheric Physics with the aid of the hybrid programming technology. Three long-term climate simulations were conducted: a control experiment (CNTR, using the Canuto2001 scheme), a K-Profile-Parameterization (KPP) experiment, and a DLVMP sensitivity experiment. Results show that DLVMP inherits the biases of KPP but exhibits improvements in simulating equatorial subsurface temperature climatology due to the inclusion of observations over the equator subsurface. Compared with the Canuto2001 scheme, both DLVMP and KPP underestimate surface mixing and overestimate subsurface mixing in the equatorial Pacific, leading to significant differences in temperature, salinity, mixed-layer depth, Atlantic Meridional Overturning Circulation (AMOC), and subtropical cell (STC) among the experiments. Additionally, the underestimated surface vertical mixing weakens the mid-latitude ventilation process, resulting in an upward shift and intensification of the STC in the DLVMP experiment, along with a significant cold temperature bias near the thermocline. This study highlights the dual challenges of training dataset size and quality for deep learning-based parameterizations, providing key insights for future improvements.