Relating pore-space gas hydrate saturation to sonic velocity data is important for remotely estimating gas hydrate concentration in sediment. In the present study, sonic velocities of gas hydrate–bearing sands are modeled using a three-phase Biot-type theory in which sand, gas hydrate, and pore fluid form three homogeneous, interwoven frameworks. This theory is developed using well log compressional and shear wave velocity data from the Mallik 5L-38 permafrost gas hydrate research well in Canada and applied to well log data from hydrate-bearing sands in the Alaskan permafrost, Gulf of Mexico, and northern Cascadia margin. Velocity-based gas hydrate saturation estimates are in good agreement with Nuclear Magneto Resonance and resistivity log estimates over the complete range of observed gas hydrate saturations.