Globally, soils contain a massive pool of carbon that is climate sensitive and offers the possibility of both mitigating and exacerbating climate change. Models such as the Canadian Land Surface Scheme including Biogeochemical Cycles (CLASSIC) simulate process-based soil carbon dynamics to understand their future behaviour. Evaluating the simulated soil carbon is not trivial as soil carbon is an inherently local-scale quantity with slow rates of change. Radiocarbon can be a useful tracer for assessing and refining model simulated soil carbon turnover. To benefit from this, we integrated a radiocarbon tracer into CLASSIC. We used atmospheric radiocarbon concentrations as a model input and compared the simulated radiocarbon distribution with observation-based reference datasets. We evaluated three versions of CLASSIC. The first is the default CLASSIC that uses bulk soil carbon pools. The second version uses a new soil carbon scheme that tracks and allows movement of carbon between model soil layers due to bio- and cryo-turbation as well as reduced respiration at depth using parameter values derived from the literature. The final version takes the second version and optimizes the parameters of the soil carbon scheme against soil carbon and soil respiration observations in a Bayesian framework. Our results indicate that default CLASSIC has too rapid a turnover of soil carbon compared to observations. The new model versions perform better, but we suggest that adding soil radiocarbon constraints to the parameter optimization would further constrain the parameter estimation and reduce the risk of parameter equifinality, further improving CLASSIC’s soil carbon projections.