Abstract Gas cyclones have many industrial applications in solid–gas separation. The vortex finder is an essential part in gas cyclones where the shape and diameter highly affect the cyclone performance. Many optimization studies have been conducted to optimize the cylindrical vortex finder diameter. This study introduces a new vortex finder shape optimized for minimum pressure drop using the discrete adjoint method. The new optimum cyclone will save 66% from the driving power needed for the Stairmand cyclone. To efficiently perform the grid independence study for the new cyclone, a new framework using the adjoint solver and the grid convergence index is proposed and tested. The proposed framework relies on local mesh adaptation instead of the global mesh refinement approach. A comparison of numerical simulation of the new cyclone and the Stairmand cyclone confirms the superior performance of the new vortex finder shape for the pressure drop and the cut-off diameter. The results of this study open a new era gas cyclones geometry optimization by using the adjoint method instead of the traditional surrogate based optimization technique. Moreover, the computational costs for the grid independence studies will be reduced via the application of the adjoint methods.