The thermal conductivity of granular planetary regolith is strongly dependent on the porosity, or packing density, of the regolith particles. However, existing models for regolith thermal conductivity predict different dependencies on porosity. Here, we use a full-field model of planetary regolith to study the relationship between regolith radiative thermal conductivity, porosity, and the particle non-isothermality. The model approximates regolith as regular and random packings of spherical particles in a 3D finite element mesh framework. Our model results, which are in good agreement with previous numerical and experimental datasets, show that random packings have a consistently higher radiative thermal conductivity than ordered packings. From our random packing results, we present a new empirical model relating regolith thermal conductivity, porosity, temperature, particle size, and the thermal conductivity of individual particles. This model shows that regolith particle size predictions from thermal inertia are largely independent of assumptions of regolith porosity, except for when the non-isothermality effect is large, as is the case when the regolith is particularly coarse and/or is composed of low thermal conductivity material.