Abstract: In order to establish prediction models for the routine smoke components released from demi-slim cigarettes based on their material parameters, cigarette samples with different material parameters were designed by a central composite design combined with an orthogonal design. Seven prediction equations for the releases of TPM (total particulate matter), tar, nicotine, CO, moisture in mainstream smoke and the open draw resistance and total ventilation rate of the demi-slim cigarettes were established by using linear regression and stepwise regression methods. The optimal models were selected according to the statistical principle of minimum root mean squares error of cross validation (RMSECV) and maximum linearly dependent coefficients between the predicted and measured values. The models were verified with commercial demi-slim cigarettes. The results showed that the average prediction relative deviations of TPM, tar, nicotine, CO and moisture releases, the open draw resistance and the total ventilation rate were 4.0%, 2.1%, 6.0%, 4.5%, 8.3%, 6.6% and 3.2%, respectively, which indicated that the models had good prediction accuracies and suitable applicability for demi-slim cigarette designs with different tobacco blends and physical material parameters.