End-to-end available bandwidth is an important factor of the performance of network, most research works focus on the self-congestion methods by far. But the self-congestion methods may induce congestion on the end-to-end path and then influence the traffic on the path seriously. In this paper, a model-based method is proposed to avoid the drawback of the self-congestion methods. Firstly, it treats Internet as a queueing system which has a single server and serves two types of traffic: the probing traffic and the background traffic, and then builds a relation between the load of the end-to-end path and the coefficient of variation of the interdeparture time of the probing traffic by analyzing the departure process of the probing traffic. Finally, the end-to-end available bandwidth is obtained from this relation. Compared with the self-congestion methods, the proposed method will not induce congestion and is more accurate.