--- title: "NVIDIA MedTech Open Models — Foundation Models for Medical AI" description: "Open-source foundation models for 3D medical image segmentation, synthetic data generation, clinical reasoning, and physical AI — runnable by coding agents via Medical AI Skills. 11 models across CT, MRI, X-ray, ultrasound, and video. Engineering and research use only." canonical: https://nvidia-medtech.github.io/ audience: [engineer, researcher, clinician] last_updated: 2026-06-10 source: index.html --- Open Source # MedTech Open Models Foundation models for 3D medical image segmentation, synthetic data generation, clinical reasoning, and physical AI — now runnable by your coding agent. [Explore Models](#physical-ai) [Run with an agent](#agent-skills) 3 Physical AI 8 Medical Imaging [ 12 Agent Skills ](#agent-skills) ## Physical AI 3 models built on Cosmos and GR00T for vision-language-action control, surgical world modeling, and sim-to-real transfer via Isaac and Holoscan ### [GR00T-H](/models/gr00t-h) Healthcare Physical AI VLA Non-Commercial Vision-language-action model for multi-embodiment surgical and clinical robot control across 16 platforms. Multi [HF](https://huggingface.co/nvidia/GR00T-H) [GitHub](https://github.com/NVIDIA-Medtech/GR00T-H) [Paper](https://arxiv.org/abs/2503.14734) ### [Cosmos-H-Surgical](/models/cosmos-h-surgical) Surgical World Model Non-Commercial Predict future surgical states and transfer simulation to photorealistic video for scalable synthetic data generation. Video [HF](https://huggingface.co/nvidia/Cosmos-H-Surgical) [GitHub](https://github.com/NVIDIA-Medtech/Cosmos-H-Surgical) [Paper](https://arxiv.org/abs/2512.23162) ### [Cosmos-H-Surgical-Simulator](/models/cosmos-h-surgical-simulator) Action-Conditioned Simulation Commercial Predict surgical outcomes from robot actions for in-silico policy evaluation and synthetic data generation. Video [HF](https://huggingface.co/nvidia/Cosmos-H-Surgical-Simulator) [GitHub](https://github.com/NVIDIA-Medtech/Cosmos-H-Surgical-Simulator) ## Medical Imaging 8 models spanning MONAI-powered segmentation, raw signal reconstruction, synthetic volume generation, and clinical VLM reasoning across CT, MRI, X-ray, and ultrasound ### [NV-Segment-CT](/models/nv-segment-ct) 132-Class CT Segmentation Commercial Automatic and interactive segmentation of 132 anatomical structures in CT images with point-click refinement. CT [HF](https://huggingface.co/nvidia/NV-Segment-CT) [GitHub](https://github.com/NVIDIA-Medtech/NV-Segment-CTMR) [Paper](https://arxiv.org/abs/2406.05285) [Run with an agent](/models/nv-segment-ct) ### [NV-Segment-CTMR](/models/nv-segment-ctmr) 345+ Class CT & MRI Segmentation Non-Commercial Automatic segmentation of 345+ anatomical structures across CT body, MRI body, and MRI brain modalities. CT, MRI [HF](https://huggingface.co/nvidia/NV-Segment-CTMR) [GitHub](https://github.com/NVIDIA-Medtech/NV-Segment-CTMR) [Paper](https://arxiv.org/abs/2406.05285) [Run with an agent](/models/nv-segment-ctmr) ### [NV-Raw2Insights-MRI](/models/nv-raw2insights-mri) Raw MRI Signal Reconstruction Commercial Transform raw MRI signals into high-quality clinical images with accelerated reconstruction and motion correction. MRI [HF](https://huggingface.co/nvidia/NV-Raw2Insights-MRI) [GitHub](https://github.com/NVIDIA-Medtech/NV-Raw2insights-MRI) ### [NV-Raw2Insights-US](/models/nv-raw2insights-us) Raw Ultrasound Processing Non-Commercial Translate raw ultrasound pressure measurements into clinical-quality images with physics-informed AI. Ultrasound [HF](https://huggingface.co/nvidia/NV-Raw2Insights-US) [GitHub](https://github.com/NVIDIA-Medtech/NV-Raw2insights-US) ### [NV-Generate-CT](/models/nv-generate-ct) 3D CT Synthesis Commercial Generate high-resolution synthetic CT volumes up to 512x512x768 with controllable anatomy and paired segmentation masks. CT [HF](https://huggingface.co/nvidia/NV-Generate-CT) [GitHub](https://github.com/NVIDIA-Medtech/NV-Generate-CTMR) [Paper](https://arxiv.org/abs/2508.05772) [Run with an agent](/models/nv-generate-ct) ### [NV-Generate-MR](/models/nv-generate-mr) 3D MRI Synthesis Non-Commercial Generate synthetic MRI volumes with Rectified Flow for 33x faster inference. Brain, abdomen, breast, and prostate coverage. MRI [HF](https://huggingface.co/nvidia/NV-Generate-MR) [GitHub](https://github.com/NVIDIA-Medtech/NV-Generate-CTMR) [Paper](https://arxiv.org/abs/2508.05772) [Run with an agent](/models/nv-generate-mr) ### [NV-Generate-MR-Brain](/models/nv-generate-mr-brain) 3D Brain MRI Synthesis Commercial Generate synthetic brain MRI across T1w, T2w, FLAIR, and SWI contrasts with cross-sequence synthesis via ControlNet. MRI [HF](https://huggingface.co/nvidia/NV-Generate-MR-Brain) [GitHub](https://github.com/NVIDIA-Medtech/NV-Generate-CTMR) [Paper](https://arxiv.org/abs/2508.05772) [Run with an agent](/models/nv-generate-mr-brain) ### [NV-Reason-CXR](/models/nv-reason-cxr) Chest X-Ray Clinical VLM Non-Commercial Vision-language model for chest X-ray interpretation with step-by-step chain-of-thought clinical reasoning. X-Ray [HF](https://huggingface.co/nvidia/NV-Reason-CXR-3B) [GitHub](https://github.com/NVIDIA-Medtech/NV-Reason-CXR) [Paper](https://arxiv.org/abs/2510.23968) [Run with an agent](/models/nv-reason-cxr) Agent Skills ## Run these models with your coding agent **12 agent skills** cover the imaging models — a run skill for the segmentation, generation, and chest-X-ray models, finetune skills, and 3 DICOM utilities. Install one and Claude Code, Cursor, or Codex runs the model correctly on the first try. Engineering and research use only. 79/81 valid first-shot commands with the skill 0/81 valid first-shot commands with the README alone Across 9 model skills and 3 agent backends — an engineering-reproducibility result. [Methodology](https://github.com/NVIDIA-Medtech/medical-AI-skills/tree/dev#trust-and-evidence). install a skill Copy ```bash npx skills add NVIDIA-Medtech/medical-AI-skills ``` Open the Agent Skills spec at [agentskills.io](https://agentskills.io), or browse all on [GitHub](https://github.com/NVIDIA-Medtech/medical-AI-skills). ### Run a model 6 [NV-Segment-CT](/models/nv-segment-ct) [NV-Segment-CTMR](/models/nv-segment-ctmr) [NV-Generate-CT](/models/nv-generate-ct) [NV-Generate-MR](/models/nv-generate-mr) [NV-Generate-MR-Brain](/models/nv-generate-mr-brain) [NV-Reason-CXR](/models/nv-reason-cxr) ### Finetune a model 3 [NV-Segment-CT](/models/nv-segment-ct) [NV-Generate-CT](/models/nv-generate-ct) [NV-Generate-MR-Brain](/models/nv-generate-mr-brain) ### DICOM utilities 3 [DICOM Metadata Extract](/skills/dicom-metadata-extract) [DICOM Series Preflight](/skills/dicom-series-preflight) [DICOM Series To Volume](/skills/dicom-series-to-volume) ## Why MedTech Open Models Built for medical imaging researchers and developers 01 ### Start from Strong Foundations Foundation models pre-trained on large-scale medical datasets. Fine-tune on your data instead of training from scratch. 02 ### From 2D to 4D CT, MRI, X-ray, ultrasound, and surgical video in one collection. Static images through dynamic physiology. 03 ### Synthetic Data Generation Create synthetic volumes to augment training, simulate rare pathologies, and enable privacy-preserving data sharing across institutions. 04 ### Open Weights Download from HuggingFace and run locally. No API keys, no cloud dependency. Apache 2.0 source with model weights for commercial and research use. 05 ### Run with Your Agent Agent skills wrap each model so Claude Code, Cursor, or Codex calls it correctly — right flags, right inputs — from a plain-language prompt. 06 ### Research to Deployment Every model includes pre-trained weights, example configs, and documentation. Go from paper to inference with minimal setup.