Online social networks play a crucial role in spreading information at a very large scale. Modeling information propagation on social networks has been attracting a lot of attention from researchers. However, none of the data sets used in past works are made available to the research community, while they would be very useful for comparative studies. In this paper, we detail a collection of tweets composed of five data sets for a total of 18 million tweets that we release, and which is designed to evaluate methods on modeling the information spread, in the case of general information and brands marketing information. In addition to tweet IDs and a script to retrieve the whole tweet in JSON from the Twitter API, we release the values of the 29 extracted features for these data sets. These features consist of user based, content based and temporal based features. Finally, we provide the results of information diffusion prediction models (80% accuracy) which could serve as strong baselines for this research topic.