GRUtopia is a simulated interactive 3D society designed for various robots, aiming to advance Embodied AI research. The platform features GRScenes, a large-scale scene dataset with 100,000 interactive, finely annotated scenes covering 89 functional categories, enabling city-scale environments. GRResidents, an NPC system driven by Large Language Models (LLMs), simulates social interactions, task generation, and assignment. GRBench, a benchmark, supports various robots, focusing on legged robots and tasks like Object Loco-Navigation, Social Loco-Navigation, and Loco-Manipulation. GRUtopia addresses the challenge of data scarcity in Embodied AI by providing a comprehensive, realistic simulation environment. The platform includes control APIs for various robots, enabling high-level task planning. GRBench evaluates robot agents' capabilities through three benchmarks, assessing navigation, social interaction, and manipulation. The platform supports extensive experiments, demonstrating the effectiveness of LLM and VLM-driven baselines. GRUtopia offers a realistic, scalable environment for Embodied AI research, with continuous improvements in scene diversity and task complexity. The platform is available at https://github.com/OpenRobotLab/GRUtopia.GRUtopia is a simulated interactive 3D society designed for various robots, aiming to advance Embodied AI research. The platform features GRScenes, a large-scale scene dataset with 100,000 interactive, finely annotated scenes covering 89 functional categories, enabling city-scale environments. GRResidents, an NPC system driven by Large Language Models (LLMs), simulates social interactions, task generation, and assignment. GRBench, a benchmark, supports various robots, focusing on legged robots and tasks like Object Loco-Navigation, Social Loco-Navigation, and Loco-Manipulation. GRUtopia addresses the challenge of data scarcity in Embodied AI by providing a comprehensive, realistic simulation environment. The platform includes control APIs for various robots, enabling high-level task planning. GRBench evaluates robot agents' capabilities through three benchmarks, assessing navigation, social interaction, and manipulation. The platform supports extensive experiments, demonstrating the effectiveness of LLM and VLM-driven baselines. GRUtopia offers a realistic, scalable environment for Embodied AI research, with continuous improvements in scene diversity and task complexity. The platform is available at https://github.com/OpenRobotLab/GRUtopia.