The embodiment of the invention provides a model training method and device, a computer device and a computer readable storage medium. The method comprises the steps of obtaining a training corpus through a first preset mode; segmenting the corpus according to preset conditions to obtain a plurality of data blocks; inputting the data blocks into the corresponding sub-models according to a preset corresponding relationship to train each sub-model to obtain trained sub-models; and synthesizing the trained sub-models according to a second preset mode to obtain a synthetic model. According to theembodiment of the invention, at the model training, the method is based on a concept of combining parallel and serial, a corpus is divided into data blocks; and the data blocks are inputted into the corresponding sub-models according to preset settings, each sub-model is trained in a parallel mode, and the serial combined calculation is carried out through a plurality of fine molecule models to form a final multi-layer composite model, thereby remarkably improving the efficiency of the model training and saving the computer hardware resources.