Skip to content
@nvidia-cosmos

NVIDIA Cosmos

NVIDIA Cosmos is a world foundation model platform for accelerating the development of physical AI systems.

NVIDIA Cosmos

NVIDIA Cosmos™ is a platform purpose-built for physical AI, featuring state-of-the-art generative world foundation models (WFMs), robust guardrails, and an accelerated data processing and curation pipeline. Designed specifically for real-world systems, Cosmos enables developers to rapidly advance physical AI applications such as autonomous vehicles (AVs), robots, and video analytics AI agents.

Cosmos World Foundation Models come in three model types which can all be customized in post-training: cosmos-predict, cosmos-transfer, and cosmos-reason:

Predict Transfer Reason
Type World Generation Multi-Controlnet Reasoning VLM
Function Predict novel future frames given initial frames Transfer existing control frames into photoreal frames within a video clip Reason against frames within a video clip
Use Cases Data Generation & Policy Evaluation Data Augmentation Data Curation
Inputs Text, Image, Video Multiple Video Modalities such as RGB, Depth, Segmentation, and more. Video & Text
Outputs Video Video Text

NVIDIA Cosmos Cookbook

The Cosmos Cookbook offers developers step-by-step recipes and post-training scripts to quickly build, customize, and deploy NVIDIA’s Cosmos world foundation models for robotics and autonomous systems.

Use Cases in Physical AI Development

Our world foundation models are purpose-built to accelerate improving performance in downstream model tasks in various stages, as illustrated here in the flywheel.

Cosmos Data Flywheel

Popular repositories Loading

  1. cosmos-reason1 cosmos-reason1 Public

    Cosmos-Reason1 models understand the physical common sense and generate appropriate embodied decisions in natural language through long chain-of-thought reasoning processes.

    Python 771 65

  2. cosmos-transfer1 cosmos-transfer1 Public

    Cosmos-Transfer1 is a world-to-world transfer model designed to bridge the perceptual divide between simulated and real-world environments.

    Python 716 98

  3. cosmos-predict2 cosmos-predict2 Public

    Cosmos-Predict2 is a collection of general-purpose world foundation models for Physical AI that can be fine-tuned into customized world models for downstream applications.

    Python 654 88

  4. cosmos-predict1 cosmos-predict1 Public

    Cosmos-Predict1 is a collection of general-purpose world foundation models for Physical AI that can be fine-tuned into customized world models for downstream applications.

    Jupyter Notebook 372 75

  5. cosmos-predict2.5 cosmos-predict2.5 Public

    Cosmos-Predict2.5, the latest version of the Cosmos World Foundation Models (WFMs) family, specialized for simulating and predicting the future state of the world in the form of video.

    Python 257 23

  6. cosmos-rl cosmos-rl Public

    Cosmos-RL is a flexible and scalable Reinforcement Learning framework specialized for Physical AI applications.

    Python 204 27

Repositories

Showing 10 of 12 repositories
  • nvidia-cosmos/cosmos-dependencies’s past year of commit activity
    Shell 3 Apache-2.0 2 1 0 Updated Nov 2, 2025
  • cosmos-rl Public

    Cosmos-RL is a flexible and scalable Reinforcement Learning framework specialized for Physical AI applications.

    nvidia-cosmos/cosmos-rl’s past year of commit activity
    Python 204 Apache-2.0 27 35 9 Updated Nov 1, 2025
  • .github Public
    nvidia-cosmos/.github’s past year of commit activity
    0 1 0 0 Updated Oct 31, 2025
  • cosmos-reason1 Public

    Cosmos-Reason1 models understand the physical common sense and generate appropriate embodied decisions in natural language through long chain-of-thought reasoning processes.

    nvidia-cosmos/cosmos-reason1’s past year of commit activity
    Python 771 Apache-2.0 65 6 5 Updated Oct 30, 2025
  • cosmos-curate Public

    Cosmos-Curate is a powerful video curation system that processes, analyzes, and organizes video content using advanced AI models and distributed computing.

    nvidia-cosmos/cosmos-curate’s past year of commit activity
    Python 100 Apache-2.0 16 0 0 Updated Oct 30, 2025
  • cosmos-cookbook Public

    Post-training scripts and samples for NVIDIA Cosmos ecosystem

    nvidia-cosmos/cosmos-cookbook’s past year of commit activity
    Python 42 Apache-2.0 7 0 0 Updated Oct 30, 2025
  • cosmos-xenna Public

    Python library for building and running distributed data pipelines using Ray

    nvidia-cosmos/cosmos-xenna’s past year of commit activity
    Python 43 Apache-2.0 9 2 4 Updated Oct 31, 2025
  • cosmos-predict2.5 Public

    Cosmos-Predict2.5, the latest version of the Cosmos World Foundation Models (WFMs) family, specialized for simulating and predicting the future state of the world in the form of video.

    nvidia-cosmos/cosmos-predict2.5’s past year of commit activity
    Python 257 Apache-2.0 23 4 3 Updated Oct 30, 2025
  • cosmos-transfer1 Public

    Cosmos-Transfer1 is a world-to-world transfer model designed to bridge the perceptual divide between simulated and real-world environments.

    nvidia-cosmos/cosmos-transfer1’s past year of commit activity
    Python 716 Apache-2.0 98 10 8 Updated Oct 30, 2025
  • cosmos-predict2 Public

    Cosmos-Predict2 is a collection of general-purpose world foundation models for Physical AI that can be fine-tuned into customized world models for downstream applications.

    nvidia-cosmos/cosmos-predict2’s past year of commit activity
    Python 654 Apache-2.0 89 10 6 Updated Oct 29, 2025