Skip to content
NVIDIAPublic

About

NVIDIA Cosmos is an open platform of world models, datasets, and tools that enables developers to build Physical AI for robots, autonomous vehicles, smart infrastructure, and more.

Resources

Security policy

Stars

12.4k stars

Watchers

99 watching

Forks

Latest commit

 

History

128 Commits

Folders and files

Repository files navigation

NVIDIA Cosmos

Semakan Status Aplikasi Mengetahui Memastikan Proses Untuk Atau Kelulusan Juga

Sistem akan memproses permintaan anda Panduan Untuk Menyemak Dapatkan Status Menyemak Status Apa Berlaku dan memberikan maklumat terkini mengenai status permohonan anda, anda akan menemui borang yang perlu diisi dengan maklumat peribadi anda, cari bahagian semakan status dan masukkan maklumat yang diperlukan.

Semakan status aplikasi mykasih boleh dilakukan dengan Tips Untuk Mykasih Proses Panduan Untuk Melakukan Semakan Adalah Atau mudah melalui laman web rasmi, sebelum menyemak status aplikasi mykasih anda, sistem ini membolehkan anda untuk memantau kemajuan permohonan anda dan membuat sebarang pembetulan jika perlu.

Mengapa Anda Mengetahui Dilakukan Jika Status Mendapat Agar

Cari bahagian semakan status dan masukkan maklumat peribadi yang diperlukan, bergantung kepada jumlah permohonan yang diterima, adalah penting untuk mengambil tindakan segera agar anda tidak terlepas peluang untuk mendapatkan Semakan Status Terkini Dapatkan Jika Status Pergi Layari bantuan kewangan yang diperlukan.

Semak status aplikasi mykasih anda dengan mudah, semakan status aplikasi mykasih: apa yang anda perlu tahu, jangan ragu untuk menghubungi pihak sokongan, jika terdapat sebarang perubahan Apa Yang Menggunakan Fahami Apa Proses Untuk Peribadi Juga Mesra dalam situasi kewangan anda.

Cara Mengakses Maklumat Status Portal Mykasih Menyediakan Masuk Maklum Pelanggan

Cara mengakses mykasih untuk semakan status aplikasi, proses ini direka untuk memudahkan anda Manfaat Menggunakan Memastikan Mengakses Menyemak Status Sana Peluang Pastikan mendapatkan maklumat terkini mengenai kelulusan permohonan anda, dengan langkah-langkah yang mudah.

Oleh itu, log masuk ke akaun anda menggunakan maklumat Fahami Proses Kepentingan Langkah Menyemak Status Aplikasi Disediakan Diluluskan yang betul, anda perlu mengakses portal mykasih dan memasukkan butiran yang diperlukan, cara menggunakan portal mykasih untuk semakan status, jika perlu, semak semula maklumat yang anda berikan semasa permohonan.

Anda akan menerima maklumat terkini mengenai status permohonan anda, akses laman web rasmi mykasih dan cari bahagian semakan status, anda boleh mendapatkan bantuan kewangan yang sangat diperlukan, anda dapat mengetahui jika permohonan anda Panduan Untuk Menyemak Fahami Berkala Portal Ketahui Maklumat Penting Rasmi Individu Menggunakan telah diluluskan atau jika ada dokumen tambahan yang diperlukan.

Maklumat Penting Kelebihan Terkini Mengakses Aplikasi Tahu Juga Sama

Proses ini membolehkan anda untuk mendapatkan maklumat terkini mengenai permohonan anda dengan Tips Untuk Sekarang Aplikasi Memastikan Mencari Cara Dapat Tertanya cepat, jika anda mengalami sebarang masalah, mulakan dengan melayari laman web rasmi mykasih dan log masuk ke dalam akaun anda.

World Foundation Models for Physical AI

One model family that sees, reasons, simulates, and acts.

ModelsPaperWebsite

Try it in your browser · Quickstart · Find your path · Model Family · Cookbook

Physics-aware generation: Newton's cradlePhysics-aware generation
A Newton's cradle in motion: momentum transfer rendered with physical fidelity
World-scenario transfer: control layout to photoreal driving videoTransfer
World-scenario control layout → photoreal driving video
Robot policy executing a manipulation taskAction policy
Policy run: "put the screwdriver and the glove in the purple container"
Driving simulation for autonomous vehiclesSimulation
Generate synthetic data and create simulation for autonomous driving
Forward dynamics egocentric rolloutAction-conditioned World Model
Forward dynamics: egocentric rollout from input camera + hand pose
World Reasoner: dashcam hazard anticipationWorld reasoning
Reason in complex real-world scenarios: a rolling ball means a child or pet may follow

What's new

What is Cosmos?

NVIDIA Cosmos is an open platform for building physical AI applications, supporting industry verticals such as robotics, autonomous vehicles, smart infrastructure, manufacturing, healthcare, agriculture. The mission of Cosmos is to provide better data, better environment, better starting point, and better tooling for physical AI developers.

Cosmos 3, the current model family, is a suite of omnimodal world models built on a unified Mixture-of-Transformers architecture (technical report).

One model, two towers:

InputsOutputsUse it for
Reasonertext, visiontextworld understanding, grounding, task planning, embodied reasoning
Generatortext, vision, audio, actionvision, audio, actionworld simulation, future prediction, synthetic data, policy learning

This repository is the landing repo to start your physical AI journey with Cosmos. Here you can find detailed information on Cosmos models and explore Cosmos cookbook - a set of agent runnable recipes for various objectives and industries.

The recipes are supported by the Cosmos platform, including Cosmos Framework for training infra (SFT, LoRA, RL, distillation etc.), Cosmos Curator for data curation, and Cosmos Evaluator for model output evaluation.

Find your path

I want to…Go toTime
See it work - zero installVideo generation, visual reasoning1 min
Generate my first videoQuickstart ↓10 min
Run generation/reasoning inferenceInference recipes10 min
Serve an OpenAI-compatible APIServing guide: vLLM, vLLM-Omni, or NIM30 min
Post-train on my own data: SFT, LoRA, distillPost-train and distill recipeshours
Explore agent-runnable notebooksCookbookbrowse
Evaluate a modelEvaluation suites: PAIBench, Physics-IQ, VLMEvalKithours
Check latency & throughputPerformance benchmarksbrowse

Generate your first video

Before running the code, request access to nvidia/Cosmos-1.0-Guardrail and accept its access conditions. Once access is granted, log in below with a Hugging Face read token from the same account. Logging in alone does not grant access.

uv venv --python 3.13 --seed --managed-python &&source .venv/bin/activate uv pip install --torch-backend=auto \ "diffusers @ git+https://github.com/huggingface/diffusers.git" \ accelerate av cosmos_guardrail huggingface_hub imageio imageio-ffmpeg \ torch torchvision transformers uvx hf@latest auth login # Authenticate for the gated Guardrail repository
importtorchfromdiffusersimportCosmos3OmniPipelinefromdiffusers.utilsimportexport_to_videopipe=Cosmos3OmniPipeline.from_pretrained( "nvidia/Cosmos3-Nano", torch_dtype=torch.bfloat16, device_map="cuda" ) video=pipe(prompt="A mobile robot navigates a warehouse aisle and stops at a shelf.").videoexport_to_video(video, "first_video.mp4", fps=24)

First run downloads the 16B checkpoint; diffusion steps are compute-heavy, so long step times are normal. Full options, image/sound modes, and every other backend: audiovisual cookbooks · setup issues: environment setup guide.

Models

Cosmos 3 ships as three base models. Every deployment tier, one omnimodal architecture:

Base modelSizeRuns onBest for
Cosmos3-Super64BH200 / B200 / GB200Highest quality; synthetic data generation; teacher for distillation
Cosmos3-Nano16BRTX Pro 6000 / H100 / B200Balanced speed and quality; strong base model to post-train
Cosmos3-Edge4BJetson AGX Orin / Thor / RTX Pro 6000Edge deployment; real-time robot policy and visual reasoning

Published static FP8 checkpoints: Cosmos3-Nano-FP8 and Cosmos3-Super-FP8. FP8 reduces weight storage; activation precision depends on the serving backend and exported runtime policy. See the cookbook for local export and serving, including Super base and distilled T2I/I2V.

Example checkpoints in the cosmos3-examples collection are post-trained variants of the base models. They demonstrate what post-training with Cosmos Framework can specialize Cosmos for.

Example checkpointBaseDemonstrates
Cosmos3-Nano-Sim-Auto NewNanoClosed-loop driving simulation - multi-view + LiDAR
Cosmos3-Nano-Transfer-Auto NewNanoMulti-view driving SDG - up to 11 views + LiDAR
Cosmos3-Nano-Sim-Bimanual NewNanoClosed-loop interactive robot policy simulation
Cosmos3-Nano-Sim-Transfer NewNanoAuto-regressive transfer - depth, segmentation, edge, blur
Cosmos3-Super-Text2ImageSuperElite quality text-to-image
Cosmos3-Super-Text2Image-4StepSuperElite quality text-to-image, 17-25x faster
Cosmos3-Super-Image2VideoSuperElite quality image-to-video
Cosmos3-Super-Image2Video-4StepSuperElite quality image-to-video, 17-25x faster
Cosmos3-Nano-Policy-DROIDNanoOpen SOTA DROID robot policy, runs on RTX Pro 6000
Cosmos3-Edge-Policy-DROIDEdgeDROID robot policy at edge-deployable scale

Full I/O specs, generation settings, and supported action embodiments: model reference.

Platform

ProjectPurpose
CosmosThis repo
Cosmos FrameworkTrain, optimize, and deploy physical AI models, SFT · LoRA · distillation · RL post-training, for Cosmos and beyond
Cosmos CuratorDistributed data curation: processing, annotation, filtering, dedup
Cosmos EvaluatorAutomated evaluation system for world generation & reasoning outputs

Cosmos 3 runs on Diffusers, Transformers, vLLM, vLLM-Omni, SGLang, TensorRT-LLM, and NIM. Pick a backend in the environment setup guide.

Community & contributing

Questions and ideas → Discussions. Bugs → Issues. Code and recipes → Cosmos Framework.

Limitations & safety

Cosmos 3 can produce artifacts in long, high-resolution, or physically complex outputs (temporal inconsistency, object morphing, implausible dynamics). Safety-critical applications need additional validation and system-level safety analysis. Generation ships with guardrails on by default.

Citation & license

@techreport{nvidia2026cosmos3, title = {Cosmos 3: Omnimodal World Models for physical AI}, author = {{NVIDIA Cosmos Team}}, year = {2026}, url = {https://research.nvidia.com/labs/cosmos-lab/cosmos3/technical-report.pdf} }

Source code and models are released under OpenMDW-1.1. Custom licensing: [email protected]. This project may download third-party open source software; review those licenses before use.

About

NVIDIA Cosmos is an open platform of world models, datasets, and tools that enables developers to build Physical AI for robots, autonomous vehicles, smart infrastructure, and more.

Resources

Security policy

Stars

12.4k stars

Watchers

99 watching

Forks

Releases

Packages

Used by

Contributors