The embodiment of the invention discloses an online model training method. The method comprises the steps of obtaining a training sample from streaming data, determining an objective function of the model according to the training sample, historical model parameters and non-convex regular terms, determining current model parameters enabling the objective function to be minimum, and updating the model according to the current model parameters. In the online training process, since the non-convex regular term is adopted to replace the L1 regular term for feature screening, the penalty deviationcan be reduced, effective features can be screened out, the sparsity is guaranteed, and the generalization performance of the model is improved. The invention further provides an information pushing method. The method comprises: obtaining user feature data and content feature data, based on the pushing model obtained by the online training model method, determining the probability that a target user is interested in target information according to the user feature data, the content feature data and the pushing model, and determining whether pushing is conducted or not according to the probability that the target user is interested in. The invention further provides an online model training device and an information pushing device.