SegPoint is a novel model that leverages large language models (LLMs) to perform point cloud segmentation across various tasks, including 3D instruction segmentation, referring segmentation, semantic segmentation, and open-vocabulary semantic segmentation. The model introduces a Geometric Enhancer Module and Geometric-guided Feature Propagation to enhance the point cloud processing capabilities of LLMs. The Geometric Enhancer Module extracts geometric information from point clouds, while the Geometric-guided Feature Propagation improves the quality of feature extraction for accurate segmentation. The model also introduces a new benchmark, Instruct3D, which evaluates segmentation performance based on complex and implicit instructions. The model achieves competitive results on established benchmarks such as ScanRefer and ScanNet, while demonstrating outstanding performance on the Instruct3D dataset. SegPoint is the first model to address these varied segmentation tasks within a single framework, achieving satisfactory performance. The model's architecture includes a pre-trained point encoder, an LLM, and the two proposed modules. The model is trained end-to-end using text classification loss and segmentation mask loss. The model's performance is evaluated on multiple datasets, including Instruct3D, ScanNet, and S3DIS, demonstrating its effectiveness in various segmentation tasks. The model's results show that it outperforms existing methods in tasks requiring intricate reasoning, and it is capable of handling both single-target and multi-target segmentation scenarios. The model's ability to understand and interpret implicit instructions makes it a promising approach for real-world applications in 3D point cloud segmentation.SegPoint is a novel model that leverages large language models (LLMs) to perform point cloud segmentation across various tasks, including 3D instruction segmentation, referring segmentation, semantic segmentation, and open-vocabulary semantic segmentation. The model introduces a Geometric Enhancer Module and Geometric-guided Feature Propagation to enhance the point cloud processing capabilities of LLMs. The Geometric Enhancer Module extracts geometric information from point clouds, while the Geometric-guided Feature Propagation improves the quality of feature extraction for accurate segmentation. The model also introduces a new benchmark, Instruct3D, which evaluates segmentation performance based on complex and implicit instructions. The model achieves competitive results on established benchmarks such as ScanRefer and ScanNet, while demonstrating outstanding performance on the Instruct3D dataset. SegPoint is the first model to address these varied segmentation tasks within a single framework, achieving satisfactory performance. The model's architecture includes a pre-trained point encoder, an LLM, and the two proposed modules. The model is trained end-to-end using text classification loss and segmentation mask loss. The model's performance is evaluated on multiple datasets, including Instruct3D, ScanNet, and S3DIS, demonstrating its effectiveness in various segmentation tasks. The model's results show that it outperforms existing methods in tasks requiring intricate reasoning, and it is capable of handling both single-target and multi-target segmentation scenarios. The model's ability to understand and interpret implicit instructions makes it a promising approach for real-world applications in 3D point cloud segmentation.