In this paper, the H∞ anti-synchronization problem of switching inertial neural networks with constant leakage delays and time-varying mixed delays is addressed. A direct analysis method based on parameterized system solutions is proposed to solve the problem we studied. The advantages of this method are that the variable substitution is not required for the system model and no Lyapunov–Krasovskii functional is constructed, which simplifies the proof process. In addition, the resulting anti-synchronization conditions consisting of some simple scalar inequalities can be easily solved by business softwares, and less computational complexity is involved. It is worth noting that the problem addressed in this paper has not been studied at present, and the limitation on the parity of activation functions in the existing literatures is removed, which reduces conservativeness. Finally, the reliability of the theoretical results is verified by numerical simulations.