NVIDIA Cosmos Reason – an open, customizable, 7B-parameter reasoning vision language model (VLM) for physical AI and robotics - enables robots and vision AI agents to reason like humans, using prior knowledge, physics understanding and common sense to understand and act in the real world. This model understands space, time, and fundamental physics, and can serve as a planning model to reason what steps an embodied agent might take next. Cosmos Reason excels at navigating the long tail of diverse scenarios of the physical world with spatial-temporal understanding. Cosmos Reason is post-trained with physical common sense and embodied reasoning data with supervised fine-tuning and reinforcement learning. It uses chain-of-thought reasoning capabilities to understand world dynamics without human annotations.

• 2025-08-08: We added the cosmos-reason1-utils inference utilities package. Adds spatial-temporal reasoning inference. See Inference for example usage.
• 2025-08-1: We added support for spatial-temporal reasoning for city and industrial operations. See latest checkpoint Cosmos-Reason1-7B.
• 2025-06-11: We enhance the model’s capability on judging the physical plausibility of a video. See this tutorial for details.
• 2025-05-17: We release model weights and training data under Hugging Face.

• Cosmos-Reason1-7B

Install system dependencies:

• pkgx

  brew install pkgx || curl https://pkgx.sh | sh

• uv

  curl -LsSf https://astral.sh/uv/install.sh | sh
  source $HOME/.local/bin/env

• Hugging Face CLI

  uv tool install -U "huggingface_hub[cli]"
  hf auth login

Clone the repository:

git clone https://github.com/nvidia-cosmos/cosmos-reason1.git
cd cosmos-reason1

Minimum Requirements:

• 1 GPU with 24GB memory

Cosmos-Reason1 is included in transformers>=4.51.3.

We provide example inference scripts:

• Minimal example

  ./scripts/inference_sample.py

• Full example

  Caption the video:

  ./scripts/inference.py --prompt prompts/caption.yaml --videos assets/sample.mp4 -v

  Ask a question about the video with reasoning:

  ./scripts/inference.py --prompt prompts/question.yaml --question 'What are the potential safety hazards?' --reasoning --videos assets/sample.mp4 -v

  Temporally caption the video and save the input frames to outputs/temporal_caption_text for debugging:

  ./scripts/inference.py --prompt prompts/temporal_caption_text.yaml --videos assets/sample.mp4 --timestamp -v -o outputs/temporal_caption_text

  Configure inference by editing:

  • Prompts
  • Sampling Parameters
  • Vision Processor Config

• Video Critic
• Post-Training
  • Full example
  • Minimal Hugging Face example
  • Minimal Llava example
• Benchmark

The nvidia-cosmos/cosmos-rl repository is an async post-training framework specialized for Supervised Fine-Tuning (SFT) and Reinforcement Learning with Human Feedback (RLHF). It prioritizes performance, scalability, and fault tolerance.

To support a custom dataset format, use the minimal Hugging Face example as a template.

The Cosmos-Reason1 model is based on the Qwen2.5-VL model architecture. Useful resources:

• Repository

This project will download and install additional third-party open source software projects. Review the license terms of these open source projects before use.

NVIDIA Cosmos source code is released under the Apache 2 License.

NVIDIA Cosmos models are released under the NVIDIA Open Model License. For a custom license, please contact [email protected].