In order to solve the problem of precision optimization design of CNC machine tools, this paper proposes an optimized allocation method of machine tool tolerance parameters that takes into account the slight deformation of the machine tool. First, establish a tolerance-based geometric error prediction model, and establish a spatial motion error model based on the theory of multi-body systems(MBS); Homogeneously, perform finite element analysis(FEA) on the CNC internal cylindrical compound grinding machine to obtain the slight deformation of the machine tool, and apply the result to the optimal tolerance allocation In the constraint conditions, the final optimal allocation plan is obtained; Finally, the genetic algorithm is used to simulate and analyze the plan, and the optimal tolerance allocation result is obtained. The result shows that most of the tolerance parameters have been relaxed, which means that the machine tool’s manufacturing cost. Through experiments, it is verified that the optimized machine tool machining accuracy pass rates are 98.5%, 98.25%, and 97.85%, respectively. Therefore, the optimal allocation method of tolerances that considers small deformations proposed in this paper is effective.