Partitioning soil organic carbon (SOC) into fractions that are stable or active on a century scale is key for more accurate models of the carbon cycle. Here, we describe the second version of a machine-learning model, named PARTY_soc, which reliably predicts the proportion of the centennially stable SOC fraction at its northwestern European validation sites with Cambisols and Luvisols, the two dominant soil groups in this region, fostering modelling works of SOC dynamics.