Accurate prediction of changes in the gas permeability during variable soil-moisture conditions is a prerequisite for improved simulation and design of soil-venting systems for removal of volatile organic chemicals in polluted soils. Air permeability, k, as a function of soil air-filled porosity, ∊, was measured on intact soil cores in a sandy forest soil from four depths with very different porosities and organic matter contents. The new data, together with 16 additional data sets for undisturbed soils, were used to evaluate several predictive models for relative gas permeability. A simplified Brooks & Corey model and a Millington & Quirk model gave similar and good predictions for sandy and loamy soils. A new soil type-dependent permeability model using the Campbell soil water retention parameter, b, to describe soil type effects gave an overall better description of relative gas permeability for all 20 soils. Both the previous and new gas permeability models require a reference point measurement of k at a relatively high air-filled porosity. However, based on measurements of both k and gas diffusion coefficients on the same undisturbed sandy and loamy soils, a simple relationship between absolute gas permeability and gas diffusivity was derived from a classical tortuous-tube gas flow model. This relation was combined with the recently presented PMQ model for predicting gas diffusivity (Soil Science 162:632-640) to obtain a new, conceptual model for absolute gas permeability as a function of soil type and airfilled porosity.