A novel, generic and efficient grid point placement method (GPPM) is developed for mesoscale modelling of composite materials. In the GPPM, grid points are employed to dynamically mark the region occupied by particles and the placement region for subsequent particles. The region within the specimen suitable for generating particles in a certain size segment is provided individually and precisely to increase the probability of success, and complex particle overlap detection can be simplified as judging the grid point state recorded in an array. Therefore, the problems of particle content and computational efficiency existing in conventional modelling methods are well solved. For arbitrary particle shapes, the GPPM has a simple data structure and can directly obtain the corresponding finite element model in ABAQUS, which has a promising application in the study of composite materials. Fully-graded concrete and fibre reinforced concrete were used as examples to validate the efficiency and feasibility of the proposed method.