Cosmos
Diffusers
Safetensors
cosmos3_omni
nvidia
cosmos3
vllm
vllm-omni
sglang
sglang-diffusion
text, image, video, audio, and action generation
omnimodel
Instructions to use nvidia/Cosmos3-Nano-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Cosmos
How to use nvidia/Cosmos3-Nano-FP8 with Cosmos:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Diffusers
How to use nvidia/Cosmos3-Nano-FP8 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("nvidia/Cosmos3-Nano-FP8", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
Commit History
2 days ago
-
Regenerate Diffusers T2V example with Nano FP8 (#10)
- Regenerate Diffusers T2V example with Nano FP8 (06a98bce0f3282e4fe1a76d5291eb660a60e1174) - Regenerate Diffusers T2V example on current main with the FP8 step policy (956dc37e6dfde64edc1eabd2d9e954a39840a917) Co-authored-by: Yiming Zhao <ConstBob@users.noreply.huggingface.co>aa5afe3 -
Sync Quality Benchmark from reviewed MR !13060
Verified signatureSync only the complete Quality Benchmark section from GitLab MR !13060 at eddc7e7a1f221b6f31ec6c107e32e91d8fc31bea. Preserve all other README sections and repository files unchanged.0792870
3 days ago
-
Update sglang description; remove unused assets (#9)
- remove benchmark images; update sglang description (85a5d84f0c7d2d258134e18375fba70db8d66c7d) Co-authored-by: Mu-Ti Chung <mutic@users.noreply.huggingface.co>722946a -
Integrate reviewed Cosmos3-Nano-FP8 model card with current usage and performance
Verified signatureUse the reviewed MR overview, quality benchmarks, and training/evaluation datasets; retain current HF usage examples and performance measurements. Add the PNG relationship diagram, set release date to November 9, 2026, and correct precision and section links. Exclude internal-only sections. Existing shared subcards are unchanged.33dde26 -
docs: fix Diffusers FP8 checkpoint restore (#5)
- docs: fix Diffusers FP8 checkpoint restore (c4dc904eed8896973e9b923c35d779487f7bafd5) Co-authored-by: Yiming Zhao <ConstBob@users.noreply.huggingface.co>b3bb52d -
Update vLLM, SGLang and vLLM-Omni examples; refresh generated examples from FP8 (#7)
- Update vLLM, SGLang and vLLM-Omni examples; refresh generated examples from FP8 (2f29c6d2db0b1638fce7d238f433f1b5ecee815a) - Point Test Hardware to the Performance Benchmarking section (b32b545c401bbdb4b82747d1aa262b0a0a98fe19) - Cap huggingface_hub below 2 and note the Rust toolchain requirement for the SGLang editable install (2a81c9a5f72c3eb1d7f9a2e77805102ab0e7c916) - Drop the nonexistent [cli] extra from the huggingface_hub install lines (7ebd7ba651b47918c80c25c46cf76f980436d140) - Resolve README conflict by retaining main Test Hardware details (576bb359a29fd6f8e9b20b17897bcd19be0ae3ab) Co-authored-by: Wojciech Kutak <wkutak@users.noreply.huggingface.co>e2af96f -
Add compact links to both Cosmos3 collections
Verified signatured249ad5
6 days ago
-
Update GB300 and DGX Reasoner performance benchmarks (#6)
- Update GB300 and DGX Reasoner performance benchmarks (d2fd3122686b46fbca689150b770e4a99269cad1) - Remove GB200 from test hardware documentation (e4f5323cd88ab036e0975d75922df8b16c78e5fd) - Expand tested hardware list (58a25f20381d5ecd2c7cc84a81f9ad8f428ac7df) - List test hardware on a single line (b3344ca64d4f49dced090ac2c3d33b81f07afa00) - Remove RTX PRO 5000 from benchmark table until results are available (52898f28b974ea67b6e19d9f442c2a5e58f242ff) Co-authored-by: Sergiy Fefilatyev <fsergiy@users.noreply.huggingface.co>67dc34e
24 days ago
-
Fill GB300 Action Policy estimate using B300 performance (#4)
- Fill GB300 Action Policy estimate using B300 performance (1935d649047961ea16061e2684fc5c0eaf7a5888) Co-authored-by: Sergiy Fefilatyev <fsergiy@users.noreply.huggingface.co>313dbe5
29 days ago
-
Consolidate Cosmos3-Nano-FP8 performance benchmarks (#3)
- Consolidate Cosmos3-Nano-FP8 performance benchmarks (58e34ad612b217682ecba373fe26a2e5b6fbc55c) Co-authored-by: Sergiy Fefilatyev <fsergiy@users.noreply.huggingface.co>6026730
30 days ago
-
Add Cosmos3 Nano FP8 snapshot (#1)
- Track Nano FP8 tokenizer and example assets (738e02cba84a52df5b80719cb2374cb522e0efd5) - Add Cosmos3 Nano FP8 snapshot files (13a8e400db8843a7686d3ffee48fbe2f499b1de9) Co-authored-by: Wojciech Kutak <wkutak@users.noreply.huggingface.co>72bd99d
-
Fix model card image and video links to original model assets
Verified signaturece38b2b -
Add FP8 model card based on nvidia/Cosmos3-Nano
Verified signature6af2039 -
initial commit
Verified signaturecd771d0