Zhengyi/LLaMA-Mesh

LLaMA-Mesh, developed by Zhengyi Wang et al. with base weights from Meta and fine-tuned by Nvidia, is an 8-billion parameter language model with a 32768-token context length. It unifies text and 3D mesh generation by representing mesh data as plain text, enabling conversational 3D generation and understanding. This model excels at generating 3D meshes from text prompts and interpreting 3D meshes, achieving quality comparable to models trained from scratch while maintaining strong text generation performance.