Molecular dynamics (MD) simulations are among the most widely used computational tools in chemistry and materials science, enabling the investigation of material properties and providing detailed insight into the time evolution of atomic systems across multiple length and time scales. The accuracy and efficiency of MD simulations depend critically on reliable force evaluations to govern atomic motion. Machine-learning (ML) approaches have become essential for constructing efficient and accurate force fields for atomistic simulations. However, high-accuracy ML models for materials typically rely on thousands to hundreds of thousands of trainable parameters and often scale poorly with increasing system size. Recent developments have extended this paradigm by incorporating quantum computational techniques, using variational quantum machine-learning (QML) models to learn potential energy surfaces and atomic forces from ab initio reference data. To date, existing QML force-field studies have largely been restricted to small molecular systems with limited chemical complexity and modest atom counts. In this work, we present a resource-efficient hybrid quantum–classical force-field framework with fewer than 100 trainable parameters that scales to chemically complex atomistic systems comprising 80–130 atoms and multiple atomic species. Despite its extremely compact parameterization, the model achieves energy prediction errors on the order of meV/atom, sufficient to drive stable molecular-dynamics simulations. By combining descriptor compression, amplitude encoding, and shallow variational quantum circuits, the proposed framework decouples model complexity from system size and establishes a practical route toward QML-driven atomistic dynamics. The approach is demonstrated on energy-relevant materials, including lithium-intercalated metal–organic frameworks and solid-state electrolytes, highlighting its potential for scalable simulations of complex materials systems.