Staggered-grid finite-difference (SGFD) approaches are universally applied to discretize different seismic-wave equations during wavefield extrapolation. However, the traditional SGFDs may encounter numerical dispersion error and instability owing to the limited approximation accuracy. To increase the simulated accuracy, we develop an optimized SGFD with high-order accuracy based on the orthogonal-octahedral operator for 3D scalar-wave modeling. Compared with the standard orthogonal-octahedral approach, the modified approach has smaller computing cost because we reduce the SGFD stencil. In addition, the corresponding time-space domain dispersion relation is beneficial to generate the least-square-based optimized high-order SGFD coefficients. Dispersion and stability comparsions show that the developed algorithm has better performance than the classical methods. Several simulated experiments verify that the proposed scheme can significantly suppress numerical dispersion in time and space domain and effectively improve the simulated accuracy and efficiency. In conclusion, the developed scheme can provide a reliable wavefield extrapolation tool for seismic imaging and inversion.