This paper presents a new model for the prediction of the compressibility factor Z = PV/RT of natural gases, either dry or condensate gases. The only requirement is the knowledge of a lumped composition. The model is relevant for temperatures and pressures that both fit petroleum engineering surface design and reservoir simulation. Predictions are expected to be within 1.7% or less for compositions including the most usual methane rich gases, but also the few cases where the light hydrocarbons are replaced by major amounts of mineral gases such as N-2, CO2, and H2S (special reservoirs or reinjection processes). This model was fitted on 207 natural gas data and is compared to the Lee-Kesler (LK) equation on a set of 220 natural gas data. For natural gas, the new model better predicts the compensations between the mixture second virial matrix coefficient contributions. As a consequence, below 500 bar the two models give similar precision accuracy except for the gas compositions where LK poorly returns the mixture mean second virial coefficient. In this latter case, the new model is much better.