Abstract Substantial amounts of carbon have been known to be stored in the miombo woodlands, which are the most widespread dry forests in the African biosphere. Efforts to get precise estimates of carbon stock in the miombo woodlands are limited by a general lack of models for estimating carbon stock. This study aimed at developing species-specific allometric models for estimating aboveground carbon stock (AGC) in Copperbelt miombo woodlands of Zambia. A total of 60 individual trees representing 4 tree species were destructively harvested for carbon analysis by Ash method. Allometric models relating diameter at breast height (DBH), total tree height (H) and basic wood density (ρ) to the AGC in form of Log10 (AGC) = a + b (Log10X1) + c (Log10X2) + d (Log10X3) were developed using Ordinary Least Squares (OLS) estimation method. Models with DBH alone performed better than those with DBH along with H and/or ρ. Evaluation of model performance was done by using R2 (Coefficient of determination), AIC (Akaike Information Criterion), Root Mean Square Error (RMSE) and the Mean Percentage Error (MPE%). Computation of MPE%, R2 and RMSE was based on the results from the k-fold cross validation technique. R2, RMSE and MPE% values for the selected models ranged from 77.3% to 98.4%, 0.065 to 0.169 and −6.09 to 1.35% respectively. Model performance and evaluation of other previously developed general and species-specific models in our dataset depicts that the models in the present study are acceptable to be applied to species in areas with similar climatic conditions. The models developed will serve as an important tool for predicting and monitoring carbon pool sizes in long-term studies. They can also be used to establish conservative carbon stock schemes required to determine avoided emissions in performance-based payment schemes.