Environmental context It is useful to know the concentration of ‘labile’, or chemically active, metal in soils because it can be used to predict metal solubility and environmental impact. Several methods for extracting the labile metal from soils have been proposed, and here we have tested two of these to see how well the resulting data can be used to model metal solubility. Such mixed approaches can be applied to different soil types with the potential to model metal solubility over large areas. Rationale Predicting terrestrial metal dynamics requires modelling of metal solubility in soils. Here, we test the ability of two geochemical speciation models that differ in complexity and data requirements (WHAM/Model VII and POSSMs), to predict metal solubility across a broad range of soil properties, using differing estimates of the labile soil metal concentration. Methodology Using a dataset of UK soils, we characterised basic properties including pH and the concentrations of humic substances, mineral oxides and metals. We estimated labile metal by extraction with 0.05 mol L−1 Na2H2EDTA and by multi-element isotopic dilution (E-value). Dissolved concentrations of Ni, Cu, Zn, Cd and Pb were estimated in 0.01 mol L−1 Ca(NO3)2 soil suspensions using the total metal ({M}total), the EDTA-extracted pool ({M}EDTA) and the E-value ({M}E) as alternative estimates of the chemically reactive metal concentration. Results Concentrations of {M}EDTA were highly correlated with values of {M}E, although some systematic overestimation was seen. Both WHAM/Model VII and POSSMs provided reasonable predictions when {M}EDTA or {M}E were used as input. WHAM/Model VII predictions were improved by fixing soil humic acid to a constant proportion of the soil organic matter, instead of the measured humic and fulvic acid concentrations. Discussion This work provides further evidence for the usefulness of speciation modelling for predicting soil metal solubility, and for the usefulness of EDTA-extracted metal as a surrogate for the labile metal pool. Predictions may be improved by more robust characterisation of the soil and porewater humic substance content and quality.