The field of human mobility modeling is witnessing significant advancements, driven by the integration of large language models (LLMs) and foundation models. Researchers are exploring innovative approaches to leverage LLMs' semantic reasoning capabilities to improve the modeling of human movement patterns. This has led to the development of new frameworks that can generate physically plausible mobility trajectories and capture patterns of normalcy in human movement. Notably, the use of self-supervised learning and bi-directional Transformers is enabling the learning of rich semantic correlations without manual labels. Furthermore, the creation of benchmarks that simulate real-world, ego-centric perception is facilitating the development of more robust trajectory forecasting systems. The incorporation of textual descriptions of external events is also enhancing dynamic urban mobility prediction. Overall, these advancements are paving the way for more comprehensive, interpretable, and powerful modeling of human mobility. Noteworthy papers include: MoveFM-R, which pioneers a new paradigm for mobility foundation models by synthesizing statistical power with deep semantic understanding. GPS-MTM, which establishes a robust foundation model for trajectory analytics by capturing patterns of normalcy in human movement. EgoTraj-Bench, which provides a critical foundation for developing trajectory forecasting systems resilient to real-world, ego-centric perception challenges. SeMob, which achieves significant reductions in prediction errors by leveraging LLM-powered semantic synthesis for dynamic mobility prediction.