MeshXL is a neural coordinate field (NeurCF) based generative model for 3D mesh generation. It represents 3D meshes as sequences of coordinates using implicit neural embeddings, enabling auto-regressive generation. The model is trained on a large-scale dataset of 3D meshes, including ShapeNet, 3D-FUTURE, Objaverse, and Objaverse-XL, resulting in over 2.5 million meshes. MeshXL is a family of pre-trained auto-regressive models that can generate high-quality 3D meshes and serve as foundation models for downstream tasks. The model uses a pre-defined ordering strategy to facilitate sequence modeling and employs decoder-only transformers for generation. Experiments show that MeshXL outperforms existing methods in terms of quality and diversity, and can generate high-quality 3D meshes with a wide range of parameters. The model is also capable of generating 3D meshes from images and text, and can be used for texture generation. MeshXL demonstrates strong performance in various tasks, including mesh completion, texture generation, and 3D mesh generation from text and images. The model is scalable and can be trained on large-scale data to improve performance. The results show that MeshXL is a promising approach for 3D mesh generation and can serve as a foundation model for various applications.MeshXL is a neural coordinate field (NeurCF) based generative model for 3D mesh generation. It represents 3D meshes as sequences of coordinates using implicit neural embeddings, enabling auto-regressive generation. The model is trained on a large-scale dataset of 3D meshes, including ShapeNet, 3D-FUTURE, Objaverse, and Objaverse-XL, resulting in over 2.5 million meshes. MeshXL is a family of pre-trained auto-regressive models that can generate high-quality 3D meshes and serve as foundation models for downstream tasks. The model uses a pre-defined ordering strategy to facilitate sequence modeling and employs decoder-only transformers for generation. Experiments show that MeshXL outperforms existing methods in terms of quality and diversity, and can generate high-quality 3D meshes with a wide range of parameters. The model is also capable of generating 3D meshes from images and text, and can be used for texture generation. MeshXL demonstrates strong performance in various tasks, including mesh completion, texture generation, and 3D mesh generation from text and images. The model is scalable and can be trained on large-scale data to improve performance. The results show that MeshXL is a promising approach for 3D mesh generation and can serve as a foundation model for various applications.