Earthquakes have been a growing concern in the state of Oklahoma in the last several years and as a result, prediction of the ground motion, understanding of subsurface geological structure, and the estimation of accurate earthquake location are of utmost importance. This entails using a high-resolution velocity model with both lateral and vertical variations. The 3D S-wave velocity model in the entire state is determined using ambient noise seismic interferometry and tomography. Passive seismic data was acquired from broadband stations in multiple networks over eight years (2009-2016). We use seismic interferometry to extract wavefields between each possible pair of stations, for all components, and for each year. Then we estimate the dispersion curve of surface waves, such as Rayleigh and Love waves, and apply seismic tomography for building 2D velocity maps at each period. Finally, we convert the velocities into 3D S-wave velocities down to about 20 km depth based on the 1D surface-wave inversion. The velocity model clearly shows slower regions associated with basins of Oklahoma and faster regions associated with shallow basement in the northeast and various uplifts throughout the state.