This letter presents an evaluation of using a hybrid neural network potential (NNP) and molecular mechanics (MM) methodology (NNP/MM) to enhance the accuracy of protein-ligand binding affinity predictions based on molecular dynamics simulations. The Alchemical Transfer Method (ATM) is employed to compute relative binding free energies (RBFE), and the performance is validated against established benchmarks. The study compares the NNP/MM approach with conventional MM force fields like GAFF2, finding significant improvements in both relative and absolute measures of accuracy. The NNP/MM method demonstrates better correlation coefficients and mean absolute errors (MAE) for most evaluated systems, although it is slower than conventional MM calculations due to the 1fs timestep limitation. The research highlights the potential of NNP/MM for accurate RBFE calculations but also acknowledges the need for further advancements in NNPs to expand their applicability to charged ligands and improve computational efficiency.This letter presents an evaluation of using a hybrid neural network potential (NNP) and molecular mechanics (MM) methodology (NNP/MM) to enhance the accuracy of protein-ligand binding affinity predictions based on molecular dynamics simulations. The Alchemical Transfer Method (ATM) is employed to compute relative binding free energies (RBFE), and the performance is validated against established benchmarks. The study compares the NNP/MM approach with conventional MM force fields like GAFF2, finding significant improvements in both relative and absolute measures of accuracy. The NNP/MM method demonstrates better correlation coefficients and mean absolute errors (MAE) for most evaluated systems, although it is slower than conventional MM calculations due to the 1fs timestep limitation. The research highlights the potential of NNP/MM for accurate RBFE calculations but also acknowledges the need for further advancements in NNPs to expand their applicability to charged ligands and improve computational efficiency.