The comprehensive assessment of severity degree of dam failure impact is an important part of dam risk evaluation and management. There are two major problems identified in the traditional comprehensive evaluation of dam break consequences: Constant weight set subjectively determined by human beings, and incompatibility between evaluation index value and evaluation grade standard. In order to solve the above-mentioned drawbacks, based on existing laws and regulations and industry norms, the life loss, economic loss, social impact and environmental impact are selected as comprehensive evaluation indexes of dam break consequences, considering the characteristics that the set pair analysis deals with the uncertainty of the system effectively and the advantage of extenics in dealing with the incompatibility problem, the variable weight set pair–extenics coupling model is established, in which the variable weight theory is used to modify the constant weight. In this model, the connection membership degree of extenics is determined according to the domain of extenics set divided by identity, difference and opposite of set pair analysis. Furthermore, the integrated connection membership degree of extenics is calculated combined with the variable weight, and is used to determine the evaluation grade according to the principle of the maximum membership. The application to the comprehensive evaluation of dam break consequences of five reservoirs in Jiangxi Province. The results show that: 1) In the comprehensive evaluation of the consequences of dam break of five reservoirs, the exponential excitation state variable weight vector motivates the poor evaluation indexes to a certain extent and increases their weight values, highlighting the status of the poor indexes in the comprehensive evaluation, avoiding the negative effects of the poor indexes being covered by the good indexes, and emphasizing their seriousness; 2) The comprehensive evaluation grades of dam break consequences of the five reservoirs are basically consistent with the evaluation grades calculated by other evaluation models, with the highest degree of dispersion and better discrimination, which shows good practicability and provides a new method for comprehensive evaluation of dam break consequences.@en