Reversible data hiding in encrypted image (RDH-EI) is a technology that not only improving the protection of the content owner’s privacy but also can retrieve the cover image after the hidden secret data being extracted. With the widely use of three-dimensional (3D) mesh models, a novel high capacity RDH-EI for 3D mesh models with highlighted multiple most significant bit ( k -MSB) prediction is proposed in this paper. We first preprocess the coordinates of each vertex in the 3D mesh models, and use an encryption key to encrypt them without k -MSB bits to construct the encrypted version. Each vertex is considered as a unit and can be predicted by its neighbors. Then, a prediction error location map can be constructed. The prediction error location map can be further highlighted and the same operation is performed in the k -MSB of the cover vertex content. The data hider can conceal the secret data into the embeddable vertices according to the highlighted results without any additional auxiliary. The hidden secret data can be extracted and/or the cover 3D mesh models can be retrieved by the receiver. Experiments show that our proposed approach can carry a higher payload comparing to some related methods.