---
title: "NV-Reason-CT — 3D CT Reasoning VLM"
description: "Vision-language model for native 3D chest and abdominal CT with structured reports, visual QA, and step-by-step reasoning."
canonical: https://nvidia-medtech.github.io/models/nv-reason-ct.html
audience: [engineer, researcher]
last_updated: 2026-09-29
source: models/nv-reason-ct.html
---
# NV-Reason-CT

3D CT Reasoning VLM

Commercial

NV-Reason-CT is a 3D vision-language model for chest and abdominal CT that writes structured reports, answers questions, and shows step-by-step reasoning. A native 3D ViT (Primus, initialized from COLIPRI) turns each volume into 13,824 visual tokens that reach a Qwen3.5-4B language model without spatial downsampling, with 3D MRoPE preserving their positions. It was trained with SFT and GRPO on about 550,000 structured QA examples from 70,111 CT volumes.

CT [HuggingFace](https://huggingface.co/nvidia/NV-Reason-CT) [GitHub](https://github.com/NVIDIA-Medtech/NV-Reason-CT) [Paper](https://arxiv.org/abs/2609.27511) [Demo](https://huggingface.co/spaces/nvidia/nv-reason-ct)

## Agent skills

### Run it with an agent

Install the skill, then just ask your agent in plain language. `medagent.nv_reason_ct`

install skill Copy

```bash
npx skills add NVIDIA-Medtech/medical-AI-skills --skill nv-reason-ct
```

Try saying

“Write structured chest and abdominal CT reports with NV-Reason-CT for the research scan data/case\_001.nii.gz — it covers both regions.”

“Run a full chest CT reasoning analysis with NV-Reason-CT on data/chest\_ct.nii.gz and show the reasoning along with the answer.”

“Ask NV-Reason-CT 'Is a pleural effusion present in this CT? Answer only Yes or No.' about data/chest\_ct.nii.gz, with thinking turned off.”

“Run the NV-Reason-CT upstream examples/example\_1.nii.gz end to end for both chest and abdomen as an engineering check, using my authorized checkout and reviewed cached model revision. I reviewed the custom model and processor code at that revision and authorize its execution with my process's permissions. Do not change my development environment.”

“Check whether this environment is ready to run NV-Reason-CT. I have not supplied a CT volume. Do not generate test data, download model files, install packages, or change my environment.”

[Skill on GitHub →](https://github.com/NVIDIA-Medtech/medical-AI-skills/tree/dev/skills/nv-reason-ct) Validation: [Benchmark](https://github.com/NVIDIA-Medtech/medical-AI-skills/blob/dev/skills/nv-reason-ct/BENCHMARK.md)·[Evals](https://github.com/NVIDIA-Medtech/medical-AI-skills/tree/dev/skills/nv-reason-ct/evals)·[Tests](https://github.com/NVIDIA-Medtech/medical-AI-skills/tree/dev/skills/nv-reason-ct/tests)·[Validators](https://github.com/NVIDIA-Medtech/medical-AI-skills/tree/dev/skills/nv-reason-ct/validators) [How agent skills work](https://agentskills.io)

Engineering & research use only — not a medical device, not for diagnosis or clinical decision-making. Skills follow the open [Agent Skills spec](https://agentskills.io).