A dimensions probability summary model is proposed, which uses dimensions probability distributions to represent the cluster summary. Similarities from point to cluster and from cluster to cluster are defined as the similarity measurement. A hierarchical clustering algorithm based on the dimensions probability summary model is proposed. Experimental analysis shows that, the model and algorithm can produce high quality of the clustering, avoid the noise and find the outliers, and automatically determine the cluster number. The proposed algorithm is stable and reliable, and is suitable for high dimensional data clustering.