Predicting information diffusion in social networks is a hard task which can lead to interesting applications: recommending relevant information for users, choosing the best entry points in the network for the best diffusion of a given piece of information, etc. We present new models which take into account three main characteristics: the number of neighbors who have disclosed the information, the relevance of the information for each user and the willingness of users to diffuse information. After this presentation, we propose to estimate the parameters of our models and illustrate their behavior through a comparison with standard information diffusion models on a real dataset. We also propose a study of the influence maximization problem associated with these new models.