A new modelling methodology for strength and stiffness prediction of discontinuous fibre-reinforced composites (DFC) is proposed. This has been validated for both thermoplastic and thermoset, prepreg based, carbon fibre reinforced, random DFC laminates having high volume fraction, by implementing it in a commercial FE solver. The methodology involves explicit generation of internal architecture of DFC through an algorithm which is efficient (faster model generation and solution), easily customizable and scalable. It captures many of the realistic features of the DFC such as variation in volume fraction, interlacing of strands, random orientation and thickness variation of strands. Thus, the model accounts for the natural mechanical property variation, which is characteristic of random DFCs and was found to be conservative in terms of prediction of tensile strength and stiffness for all the validation cases considered. It is generic in the sense that it can be easily extended to generate preferentially aligned and hybrid DFC laminates. © 2019 Elsevier Ltd