Provided are a model parameter fusion method and apparatus, applied to a machine learning system. The machine learning system comprises at least one parameter collection group and at least one parameter distribution group, wherein each parameter collection group corresponds to at least one parameter distribution group. The present invention relates to the field of machine learning, and is used to solve the problems that performance requirements for a parameter server are high during model parameter fusion, the data transmission amount is large and a calculation resource is dynamically adjusted. The method comprises: when any parameter collection group satisfies an intra-group fusion condition, fusing model parameters of M nodes in a parameter collection group to obtain a first model parameter of the parameter collection group, wherein M is greater than or equal to the lowest number s of fusion nodes of the parameter collection group, and is less than or equal to the total number of nodes contained in the parameter collection group; and sending the first model parameter of the parameter collection group to N nodes of a parameter distribution group corresponding to the parameter collection group, wherein N is greater than or equal to one and is less than or equal to the total number of nodes contained in the parameter distribution group corresponding to the parameter collection group.