Recommendation technologies have traditionally been used in domains such as E-commerce and Web navigation to recommend resources to customers so as to help them to get the pertinent resources. Among the possible approaches is collaborative filtering that does not take into account the content of the resources: only the traces of usage of the resources are considered. State of the art models, such as sequential association-rules and Markov models, that can be used in the frame of privacy concerns, are usually studied in terms of performance, state space complexity and time complexity. Many of them have a large time complexity and require a long time to compute recommendations. However, there are domains of application of the models where recommendations may be required quickly. This paper focuses on the study of how these state of the art models can be adapted so as to be anytime. In that case recommendations can be proposed to the user whatever is the computation time available, the quality of the recommendations increases according to the computation time. We show that such models can be adapted so as to be anytime and we propose several strategies to compute recommendations iteratively. We also show that the computation time needed by these new models is not increased compared to classical ones; even so, it sometimes decreases.