The soil-gas diffusion is a primary driver of transport, reactions, emissions, and uptake of vadose zone gases, including oxygen, greenhouse gases, fumigants, and spilled volatile organics. The soil-gas diffusion coefficient, Dp, depends not only on soil moisture content, texture, and compaction but also on the local-scale variability of these. Different predictive models have been developed to estimate Dp in intact and repacked soil, but clear guidelines for model choice at a given soil state are lacking. In this study, the water-induced linear reduction (WLR) model for repacked soil is made adaptive for different soil structure conditions (repacked, intact) by introducing a media complexity factor (Cm) in the dry media term of the model. With Cm = 1, the new structure-dependent WLR (SWLR) model accurately predicted soil-gas diffusivity (Dp/Do, where Do is the gas diffusion coefficient in free air) in repacked soils containing between 0 and 54% clay. With Cm = 2.1, the SWLR model on average gave excellent predictions for 290 intact soils, performing well across soil depths, textures, and compactions (dry bulk densities). The SWLR model generally outperformed similar, simple Dp/Do models also depending solely on total and air-filled porosity. With Cm = 3, the SWLR performed well as a lower-limit Dp/Do model, which is useful in terms of predicting critical air-filled porosity for adequate soil aeration. Because the SWLR model distinguishes between and well represents both repacked and intact soil conditions, this model is recommended for use in simulations of gas diffusion and fate in the soil vadose zone, for example, as a key element in developing more accurate climate change models.