This dataset contains volumetric segmentations for a sample of Computed Tomography (CT) images from the National Lung Screening (NLST) collection, publicly available from NCI Imaging Data Commons, generated using open-source AI models. This dataset was generated as part of a project evaluating concordance among the AI models and accuracy of the segmentation. The details of the study will be published as an upcoming preprint. Segmentation results from the following models are included: TotalSegmentator 1.5 (https://github.com/wasserth/TotalSegmentator) is a multi-organ segmentation model for CT images released in September 2023, with the NLST segmentations released in IDC in 2024. It is based on the nnU-Net architecture and segments 104 anatomical structures across the entire body. The model was trained on 1,204 clinical CT scans from the University Hospital Basel, collected in 2012, 2016, and 2020. TotalSegmentator 2.6 (https://github.com/wasserth/TotalSegmentator) is an extended version released in January 2025 that supports both CT and MR images. The updated total task for CT includes 117 anatomical structures. In addition, 20 open-source tasks have been added for the targeted segmentation of specific anatomical regions and pathologies (e.g., head and neck, liver segments, vessels, implants). Across all tasks, the model can segment up to 238 CT and 85 MR structures. Auto3DSeg (https://github.com/Project-MONAI/tutorials/tree/main/auto3dseg) is an automated segmentation framework built on MONAI that constructs full 3D segmentation pipelines with minimal user input. For our analysis, we used an Auto3DSeg model that was trained on the publicly available TotalSegmentator dataset. This resulted in the segmentation of 117 anatomical structures in CT images. MultiTalent (https://github.com/MIC-DKFZ/MultiTalent), released in May 2023, is based on nnU-Netv2 and trained on 13 publicly available abdominal CT datasets, totaling 1,477 volumes. To improve generalizability, the training data included specialized datasets covering different anatomical subregions (e.g., organs, vessels, tumors) as well as the TotalSegmentator dataset. The final model segments 117 anatomical structures. MOOSE 3.0 (https://github.com/ENHANCE-PET/MOOSE), released in December 2022, is a large-scale multi-organ segmentation model trained on 50 low-dose whole-body CT scans and 34 PET/MR brain datasets. It segments 143 anatomical structures using specialized clinical submodels focused on organ systems such as the heart, lungs, digestive tract, spine, and musculature. The model builds on a 3D nnU-Net architecture and incorporates a data-centric training strategy to optimize performance across heterogeneous datasets. CADS (https://github.com/murong-xu/CADS/tree/main), released in July 2025, is an open-source multi-organ segmentation model for CT images trained on the CADS-dataset, which contains 22,022 CT scans from over 100 centers in 16 countries. Importantly, this model was trained using 7172 CT scans from the NLST collection. It segments 167 anatomical structures from head to knees using a nnU-Net–based architecture with region-specific submodels. The model was validated on 18 public datasets and a large real-world oncology cohort. The segmentations are saved in DICOM Segmentation (DICOM SEG) format. The dataset files are organized as follows: _.zip: segmentations stored matching the organization of the outputs generated by the models (i.e., one DICOM SEG per model outout) OMAS_one_SEG_per_CT.zip: segmentations produced by the CADS model (eaerlier known as "OMAS") with all model outputs stored in a single DICOM SEG MOOSE_one_SEG_per_CT.zip: segmentations produced by the MOOSE model with all model outputs stored in a single DICOM SEG Consensus_segmentation.zip: consensus segmentation (voxels labeled as the specific structure by each of the models are included)