Rhododendron species are broad-leaved evergreen woody shrubs belonging to Ericaceae, and are an important constituent of alpine and subalpine ecosystems. Rh. species are endemic to Tibetan Plateau and surrounding areas. Unlike for trees, biomass estimation models are virtually lacking for Rh. species in natural communities. Therefore, accurate measurement and modeling of biomass for alpine-subalpine Rh. species is a fundamental work for quantifying carbon functions of terrestrial ecosystems. This study aimed to develop allometric models for the estimation of biomass storage of seven Rh. species' in the alpine-subalpine region of southern Gansu Province. Investigated species included Rh. rufum, Rh. przewalskii, Rh. alophytum, Rh. oreodoxa, Rh. taibaiense, Rh. capitatum and Rh. maculiferum. A total of 312 individuals were harvested for the measurements of above- and belowground biomass. Commonly used models, such as linear, logarithmic, power-law and exponential functions were used for estimating biomass, and basal diameter (D), height (H), canopy (C), crown volume (V) and square of basal diameter×plant height (D²H) from field measurements were used as independent variables, and leaf biomass, stem biomass, aboveground biomass, root biomass and total biomass were treated as dependent variables. Among a total number of 700 models tested, significant relationships were detected between biomass components and field measurements of predictors for the seven studied woody species (all P<0.01). Using D and D²H as independent variables and the power function resulted in a relatively narrow distribution and a high median of R² values. The R² values of the selected 35 optimal models for individual species from the 700 sets varied between 0.66 and 0.99, with a median of 0.92.The models for stem biomass, leaf biomass and aboveground biomass of Rh. oreodoxa were linear functions, for each of these biomass of Rh. maculiferum were exponential functions, while the rest of the biomass models were power functions. The plant height (H) was the best independent variable for the estimation of stem biomass, leaf biomass and aboveground biomass of Rh. rufum and leaf biomass of Rh. calophytum. The results also showed that the power function with D²H as the independent variable was the best model for mixed species, but the predictive power for leaf biomass was relatively low using the mixed-species' model. The establishment of seven alpine rhododendron shrub biomass models in southern Gansu Provide can serve as an important tool for the study of the carbon sink function of shrub ecosystems in alpine regions. In addition, these models may be applied to Rh. woodland elsewhere in eastern Tibetan Plateau.