World ModelResearch

NVIDIA Cosmos

NVIDIA

World foundation model platform for physical AI, built around predictive world modeling and data processing workflows.

Code: Not listed
Weights: Available
Checkpoint: Available
License: Unverified

Weights Paper Model index A–Z

Decision summary

Best for

Future-state prediction quality

Main blocker

Hardware, latency, dependencies, and checkpoint loading are not pipeline-tested.

Next check

Verify repository and artifact licenses.

Release history

Source-backed release timing for this canonical model record.

Important releases

2025-01-07
Model series introduced
Release timing is anchored to the cited paper publication date.
Paper

Evidence profile

A visual read of adoption evidence. Scores describe catalog evidence readiness, not task performance.

Published evaluation

4/6

dimensions scored

Unknown evidence remains visible and is never treated as a zero.

Access and governanceNot scoredconf. 20

Artifact availability90conf. 90

Training and loading reproducibility70conf. 65

Training data requirements76conf. 70

Evaluation evidenceNot scoredconf. 10

Deployment readiness70conf. 55

Why these scores

The strongest decision reasons behind the evidence profile.

Top 3 adoption signals

Artifact availability

Useful

Checkpoint files were verified in the official repository metadata.

90 · confidence 90

Training and loading reproducibility

Useful

Configuration files are present; runtime loading is not tested.

70 · confidence 65

Evaluation evidence

Unknown

No structured evaluation evidence has been verified.

— · confidence 10

Data requirements

Declared loop-data needs, missing evidence, and the next checks that matter.

3 required categories

Observation / ego video

observation · depth · camera pose

required

Future state / dynamics

future state

required

Sim-real / embodiment metadata

depth · camera pose · sim-real

required

Critical gaps

Core categories are represented. Interface alignment and data quality still require verification.

Linked datasets

Signal-level links only. Verify runtime interfaces before use.

Browse datasets

Ego-1K

View synthesis for wrist/ego cameras

Useful

provisional fit 100 · confidence 60

2 matched · 0 unknown

Xperience-10M

World model pretraining

Useful

provisional fit 100 · confidence 60

2 matched · 0 unknown

Ego-Exo4D

human_demo

Useful

provisional fit 100 · confidence 60

2 matched · 0 unknown

More evidenceShow

Additional decision dimensions

Access and governanceUnknown

License or access terms are not verified.

— · c20

Training data requirementsUseful

Required signal categories are structured; exact tensor and action interfaces still need verification.

76 · c70

Deployment readinessUseful

Hardware, latency, dependencies, and checkpoint loading are not pipeline-tested.

70 · c55

Artifact facts

release_year: official_claim; 2025; curated official source year
model.required_signals: official_claim; observation · depth · camera pose · sim-real · future state; Official model documentation and curated signal mapping
model.hub: official_claim; declared; curated model hub link
model.weight_files: metadata_verified; repository: nvidia/Cosmos-1.0-Diffusion-7B-Text2World · count: 264 · sample files: Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/.no_exist/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model.safetensors · Cosmos-1.0-Guardrail/aegis/models--nvidia--Aegis-AI-Content-Safety-LlamaGuard-Defensive-1.0/snapshots/f54cb2302ee876705dc0f7df2288f442c034b2f3/adapter_model.safetensors · Cosmos-1.0-Guardrail/face_blur_filter/Resnet50_Final.pth · Cosmos-1.0-Guardrail/video_content_safety_filter/safety_filter.pt · model.pt · nemo/common.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_0_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_10_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_11_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_12_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_13_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_14_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_15_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_16_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_17_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_18_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_19_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_1_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_20_28.pt · nemo/module.decoder.layers.cross_attention.core_attention._extra_state/shard_21_28.pt · repository: nvidia/Cosmos-1.0-Autoregressive-4B · count: 10 · sample files: Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/.no_exist/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model.safetensors · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model-00001-of-00003.safetensors · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model-00002-of-00003.safetensors · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model-00003-of-00003.safetensors · Cosmos-1.0-Guardrail/aegis/models--nvidia--Aegis-AI-Content-Safety-LlamaGuard-Defensive-1.0/snapshots/f54cb2302ee876705dc0f7df2288f442c034b2f3/adapter_model.safetensors · Cosmos-1.0-Guardrail/face_blur_filter/Resnet50_Final.pth · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/snapshots/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/model.safetensors · Cosmos-1.0-Guardrail/video_content_safety_filter/safety_filter.pt · model.pt · nemo/weights/common.pt; Hugging Face model metadata siblings
model.checkpoint_files: metadata_verified; repository: nvidia/Cosmos-1.0-Diffusion-7B-Text2World · count: 9 · sample files: Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/.no_exist/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model.safetensors · Cosmos-1.0-Guardrail/aegis/models--nvidia--Aegis-AI-Content-Safety-LlamaGuard-Defensive-1.0/snapshots/f54cb2302ee876705dc0f7df2288f442c034b2f3/adapter_model.safetensors · model.pt · text_encoder/model-00001-of-00002.safetensors · text_encoder/model-00002-of-00002.safetensors · transformer/diffusion_pytorch_model-00001-of-00003.safetensors · transformer/diffusion_pytorch_model-00002-of-00003.safetensors · transformer/diffusion_pytorch_model-00003-of-00003.safetensors · vae/diffusion_pytorch_model.safetensors · repository: nvidia/Cosmos-1.0-Autoregressive-4B · count: 8 · sample files: Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/.no_exist/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model.safetensors · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model-00001-of-00003.safetensors · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model-00002-of-00003.safetensors · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/model-00003-of-00003.safetensors · Cosmos-1.0-Guardrail/aegis/models--nvidia--Aegis-AI-Content-Safety-LlamaGuard-Defensive-1.0/snapshots/f54cb2302ee876705dc0f7df2288f442c034b2f3/adapter_model.safetensors · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/snapshots/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/model.safetensors · model.pt · nemo/weights/common.pt; Hugging Face model metadata checkpoint files
model.config_files: metadata_verified; repository: nvidia/Cosmos-1.0-Diffusion-7B-Text2World · count: 13 · sample files: Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/config.json · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/generation_config.json · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/tokenizer_config.json · Cosmos-1.0-Guardrail/config.json · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/.no_exist/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/processor_config.json · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/snapshots/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/config.json · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/snapshots/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/preprocessor_config.json · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/snapshots/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/tokenizer_config.json · config.json · text_encoder/config.json · tokenizer/tokenizer_config.json · transformer/config.json · vae/config.json · repository: nvidia/Cosmos-1.0-Autoregressive-4B · count: 9 · sample files: Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/config.json · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/generation_config.json · Cosmos-1.0-Guardrail/aegis/models--meta-llama--LlamaGuard-7b/snapshots/dfcfa3409b9994a4722d44e05f82e81ea73c5106/tokenizer_config.json · Cosmos-1.0-Guardrail/config.json · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/.no_exist/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/processor_config.json · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/snapshots/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/config.json · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/snapshots/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/preprocessor_config.json · Cosmos-1.0-Guardrail/video_content_safety_filter/models--google--siglip-so400m-patch14-384/snapshots/9fdffc58afc957d1a03a25b10dba0329ab15c2a3/tokenizer_config.json · config.json; Hugging Face model metadata siblings
release_timing: metadata_verified; 2025-01-07T06:55:50.000Z; arXiv 2501.03575 published

Additional references

NVIDIA CosmosProject HF autoregressive modelModel hub

OpenBot notes

Useful anchor for the world-model side of OpenBot's Catalog and Synth story.
Dataset pages should expose geometry and sim-real signals, not only labels.

NVIDIA Cosmos

Decision summary

Release history

Model series introduced

Evidence profile

Why these scores

Artifact availability

Training and loading reproducibility

Evaluation evidence

Data requirements

Observation / ego video

Future state / dynamics

Sim-real / embodiment metadata

Critical gaps

Linked datasets

Ego-1K

Xperience-10M

Ego-Exo4D

Additional decision dimensions

Artifact facts

Additional references

OpenBot notes