Differential code bias (DCB) significantly affects the ionosphere modeling, precise positioning, and navigation applications when using code observations. With the fully completed BeiDou navigation satellite system (BDS-3), there exist various DCBs of new frequencies and types which should be handled. However, limited types of DCB products for BDS-3 are provided by the analysis institutions (e.g., Chinese Academy of Science (CAS) and German Aerospace Center (DLR)). Hence, for some DCB corrections of new frequencies, they are generally generated by complex linear combinations, which are not friendly to users and may degrade the accuracy. In this study, the estimation method of DCB for BDS-3 is introduced first. Then, the BDS-3 observations from 40 globally distributed stations are selected to estimate the DCBs, including 19 types of DCBs of new frequencies for BDS-3. Moreover, the estimated DCBs are carefully analyzed in terms of inner consistency, external consistency, and stability. For the results of inner consistency, most closure error series are within 0.2 ns, and the closure error series of each satellite fluctuate near zero and have no obvious systematic deviations. For the results of external consistency, the mean deviations of estimated DCBs of each satellite are mainly within 0.3 ns and 0.2 ns for the common types of DCB products of CAS and DLR, respectively. For the results of stability, the mean values of monthly STDs for the estimated DCBs are all smaller than 0.12 ns, which exhibits good stability. The STDs of the directly estimated DCBs are generally smaller than that of the DCB combinations of DLR and CAS. In this sense, the directly estimated DCBs for BDS-3 exhibits good performance in terms of accuracy and stability in this study, which can further provide the DCB corrections for precise positioning and navigation applications.