We developed an ensemble approach that combines three agroecosystem models to predict crop yields and changes in soil carbon across the United States. The ensemble results were more accurate and consistent compared to individual model. Ensemble result matched closely the observed yield data and soil carbon measurements while reducing the uncertainty from the individual models. This work improves our ability to track carbon change and supports carbon farming, climate action, and land management.