Accurate estimate of forest biomass is essential to understand the global carbon cycle. Existing studies on aboveground biomass (AGB) estimate conducted in karst forests present high uncertainties, and no benchmarking and validation data can be used to evaluate their accuracy. In the present study, the AGB of a karst forest in central Guizhou Province, southwestern China, was accurately estimated for the first time on the basis of a clear cutting plot. The performances of existing allometric models and average standard tree method were also tested. The accurate AGB of the karst forest was 122.81 Mg/hm². Living woody individuals with diameter at breast height (D) ≥ 1 cm and D < 1 cm and herbs were 120.00 Mg/hm², 2.56 Mg/hm², and 0.24 Mg/hm², respectively. Individuals within 10-30 cm (83.89 Mg/hm²) D classes were the major AGB contributors. The four dominant species (Itea yunnanensis, Lindera pulcherrima, Cinnamomum bodinieri and Platycarya strobilacea) with 103.03 Mg/hm² AGB accounted for 83.89% of the total forest AGB. Stem (61.04 Mg/hm²) and branch (40.56 Mg/hm²) exhibited much higher AGB than bark (11.61 Mg/hm²) and leaf (6.80 Mg/hm²). Both existing allometric models (bias: -56.10% to 84.61%) and average standard tree method (-36.43% to -5.14%) had low performances indicated by high bias when estimating AGB at species level. Finally, five new local allometric models were developed. These results provide benchmarking data and further guidance for accurate estimate of forest carbon stocks in karst geomorphologies in southwestern China.