Food insecurity in sub-Saharan Africa (SSA) is partly due to low staple crop yields, resulting from low soil fertility and low nutrient inputs. Integrated soil fertility management (ISFM), that includes the combined use of mineral and organic fertilizers, can contribute to increasing yields and sustaining soil organic carbon (SOC) in the long term. Soil-crop simulation models help assess the performance of ISFM under current and future climate. Yet, uncertainty in model predictions can be high, resulting from poor model calibration and/or inadequate model structure. Multi-model assessments help understand model uncertainty and reduce it. We compared the performance of 16 soil-crop models using data from four long-term experiments at sites in SSA with contrasting climates and soils. Each experiment had four experimental treatments: i) no exogenous inputs, ii) addition of mineral nitrogen (N), iii) organic amendments, and iv) combined mineral and organic inputs. We assessed model performance after partial and full calibration. Model ensemble accuracy was greater with full calibration than with partial calibration, and improved more for crop yields (rRMSE 53 vs 17%) than for SOC (RMSE 21 vs 12%). Uncertainty of model simulations increased over the course of the long-term experiments. Uncertainty with SOC simulations increased when organic amendments were added to the soil, whilst uncertainty with yield predictions was largest when no inputs were applied. Differences between individual models were linked to uncertainties in simulating roots and soil N dynamics. The discrepancies between models increased when organic amendments were applied. Our study is the first multi-model comparison for long-term simulations of crop yield and SOC and their feedbacks in SSA. The results highlight the need for long-term experiments that monitor roots and N dynamics and provide the corresponding data required for soil-crop model improvements and calibration.