Grazing lands play a significant role in global carbon (C) dynamics, holding substantial soil organic carbon (SOC) stocks. However, historical mismanagement (e.g., overgrazing and land-use change) has led to substantial SOC losses. Regenerative practices, such as adaptive multi-paddock (AMP) grazing, offer a promising avenue to improve soil health and help combat climate change by increasing SOC accrual, both in its particulate (POC) and mineral-associated (MAOC) organic C components. Because adaptive grazing patterns emerge from the combination of different levers such as frequency, intensity, and timing of grazing, studying AMP grazing management in experimental trials and representing it in models remains challenging. Existing ecosystem models lack the capacity to predict how different adaptive grazing levers affect SOC storage and its distribution between POC and MAOC and along the soil profile accurately. Therefore, they cannot adequately assist decision-makers in effectively optimizing adaptive practices based on SOC outcomes. Here, we address this critical gap by developing version 2.34 of the MEMS 2 model. This version advances the previous by incorporating perennial grass growth and grazing submodules to simulate grass green-up and dormancy, reserve organ dynamics, the influence of standing dead plant mass on new plant growth, grass and supplemental feed consumption by animals, and their feces and urine input to soil. Using data from grazing experiments in the southeastern United States and experimental SOC data from two conventional and three AMP grazing sites in Mississippi, we tested the capacity of MEMS 2.34 to simulate grass forage production, total SOC, POC, and MAOC dynamics to 1-m depth. Further, we manipulated grazing management levers, i.e., timing, intensity, and frequency, to do a sensitivity analysis of their effects on SOC dynamics in the long term. Our findings indicate that the model can represent bahiagrass forage production (BIAS = 9.51 g C m