Complementary aspects of tissue microstructure can be studied with diffusion‐weighted imaging (DWI). However, there is no consensus on how to design a diffusion acquisition protocol for multiple models within a clinically feasible time. The purpose of this study is to provide a flexible framework that is able to optimize the shell acquisition protocol given a set of DWI models. Eleven healthy subjects underwent an extensive DWI acquisition protocol, including 15 candidate shells, ranging from 10 to 3500 s/mm2. The proposed framework aims to determine the optimized acquisition scheme (OAS) with a data‐driven procedure minimizing the squared error of model‐estimated parameters. We tested the proposed method over five heterogeneous DWI models exploiting both low and high b‐values (i.e., diffusion tensor imaging [DTI], free water, intra‐voxel incoherent motion [IVIM], diffusion kurtosis imaging [DKI], and neurite orientation dispersion and density imaging [NODDI]). A voxel‐level and region of interest (ROI)‐level analysis was conducted over the white matter and in 48 fiber bundles, respectively. Results showed that acquiring data for the five abovementioned models via OAS requires 14 min, compared with 35 min for the joint recommended acquisition protocol. The parameters derived from the reference acquisition scheme and the OAS are comparable in terms of estimated values, noise, and tissue contrast. Furthermore, the power analysis showed that the OAS retains the potential sensitivity to group‐level differences in the parameters of interest, with the exception of the free water model. Overall, there is a linear correspondence (R2 = 0.91) between OAS and reference‐derived parameters. In conclusion, the proposed framework optimizes the shell acquisition scheme for a given set of DWI models (i.e., DTI, free water, IVIM, DKI, and NODDI), combining low and high b‐values while saving acquisition time.