Structured magnetorheological elastomers (MREs) are composite materials that exhibit complex magneto-mechanical coupling effects, including the magnetostrictive and magnetorheological effect. They consist of magnetizable particles arranged in chain-like structures embedded in a soft elastomer matrix. Since explicitly resolving the microstructure of real-world samples is infeasible, a multiscale modeling approach is required. We present a framework for the macroscale modeling of structured MREs based on physics-augmented neural networks (PANNs) [1–3], explicitly accounting for the material’s anisotropic behavior [3]. The proposed PANN macromodel satisfies fundamental physical requirements such as objectivity, material symmetry, and thermodynamic consistency [1,2]. Assuming magnetostatic conditions, the framework relies on finite-element (FE) simulations formulated via a variational principle for magneto-hyperelastic materials [4]. We treat the quasi-incompressibility of the elastomer matrix using a four-field formulation. The framework starts with data generation, in which a representative volume element (RVE) is subjected to sampled macroscopic magneto-mechanical loading paths in FE simulations. The resulting homogenized microscale quantities form a macroscale dataset used for training and testing the PANN macromodel. After training, the PANN macromodel accurately predicts magnetization, mechanical stress, and total stress within the range of the training data. Finally, the trained PANN model is employed as a constitutive model in a decoupled multiscale scheme for macroscale FE simulations. This enables the quantification of magnetostrictive and magnetorheological effects of a macroscopic sample. Referenzes: [1] H.T. Roth, P. Gebhart, K.A. Kalina, T. Wallmersperger, M. Kästner, A data-driven multiscale scheme for anisotropic finite strain magneto-elasticity, arXiv:2510.24197, 2025. [2] K.A. Kalina, P. Gebhart, J. Brummund, L. Linden, W. Sun, M. Kästner, Neural network-based multiscale modeling of finite strain magneto-elasticity with relaxed convexity criteria, CMAME 421, 2025. [3] K.A. Kalina, J. Brummund, W. Sun, M. Kästner, Neural networks meet anisotropic hyperelasticity: A framework based on generalized structure tensors and isotropic tensor functions, CMAME 437, 2025. [4] P. Gebhart, Skalenübergreifende Modellierung magneto-aktiver Polymere auf Grundlage energie-basierter Variationsprinzipien, PhD thesis, TU Dresden, 2024.