Social network services and microblog are important application modes of Web2.0, in which opinion spread is quite different from that in other cyber-media and in traditional media. In this paper, we present an opinion-spreading model based on online social network services, to study the forms and features of the spread of public opinion in social network services. The simulation results show that the model can fit the actual data from a social network site. The speed of information spread is consistent with the conclusion of six degrees of separation theory. When an opinion with strong tendency spreads in a network in which intrinsic views obey uniform distribution, the opposite view cannot exist in the stable network. In a stable network, view distribution is related to the degree of source node and the depth of backtrack, but not related to confidence limit, which is different from Deffuant model and Hegselmann-Krause model. Meanwhile, in this paper, we the also analyze the influence of the probability of spreading will, the probability of opinion change and confidence limit on relaxation time. Finally, in the paper are shown two applications of the model: the spread of rumors and the role of opinion leaders.