Objaverse is a large-scale dataset of 3D objects with over 800,000 models, including detailed descriptions, tags, and animations. It is sourced from Sketchfab, a leading 3D model marketplace, and contains a diverse range of objects, including animals, humans, vehicles, and environments. The dataset is designed to support a wide range of research in computer vision, including 3D generative modeling, instance segmentation, open-vocabulary object navigation, and robustness analysis of vision models. Objaverse addresses the lack of large-scale 3D datasets by providing a rich, diverse, and high-quality collection of 3D assets. It includes 44,000 animated objects and over 63,000 objects self-categorized as characters. The dataset also includes a wide variety of visual styles, such as 3D scans, 3D modeled objects, point clouds, and photo-realistic renderings. Objaverse is used to train models for tasks such as 3D generative modeling, instance segmentation, and open-vocabulary object navigation. It also enables the creation of benchmarks for evaluating the robustness of vision models to perspective shifts. The dataset has been shown to improve the performance of instance segmentation models and to enable the training of embodied AI agents. Objaverse provides a valuable resource for researchers in computer vision and AI, enabling new applications and research directions.Objaverse is a large-scale dataset of 3D objects with over 800,000 models, including detailed descriptions, tags, and animations. It is sourced from Sketchfab, a leading 3D model marketplace, and contains a diverse range of objects, including animals, humans, vehicles, and environments. The dataset is designed to support a wide range of research in computer vision, including 3D generative modeling, instance segmentation, open-vocabulary object navigation, and robustness analysis of vision models. Objaverse addresses the lack of large-scale 3D datasets by providing a rich, diverse, and high-quality collection of 3D assets. It includes 44,000 animated objects and over 63,000 objects self-categorized as characters. The dataset also includes a wide variety of visual styles, such as 3D scans, 3D modeled objects, point clouds, and photo-realistic renderings. Objaverse is used to train models for tasks such as 3D generative modeling, instance segmentation, and open-vocabulary object navigation. It also enables the creation of benchmarks for evaluating the robustness of vision models to perspective shifts. The dataset has been shown to improve the performance of instance segmentation models and to enable the training of embodied AI agents. Objaverse provides a valuable resource for researchers in computer vision and AI, enabling new applications and research directions.