Cosmos3-Nano is an open-weight model from Ares Realm Studios, released under other. It has 15.8B parameters and a 262,144-token context. At 16-bit it needs about 37.8 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index.
NVIDIA Cosmos™ is a world foundation model platform designed to accelerate the development of Physical AI by enabling machines to understand, simulate, and interact with the physical world across robotics, autonomous driving, and smart space environments…
Runs On
What it takes to serve Cosmos3-Nano (15.8B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 31.5 GB | 37.8 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 15.8 GB | 18.9 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 7.9 GB | 9.5 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Oct 16, 2026.
Cosmos3-Nano on every accelerator the SAVRN Index prices, at every precision
Model Card
NVIDIA Cosmos™ is a world foundation model platform designed to accelerate the development of Physical AI by enabling machines to understand, simulate, and interact with the physical world across robotics, autonomous driving, and smart space environments, including industrial and factory-scale applications. Cosmos3 is a collection of Omnimodal world models capable of generating dynamic, high-quality video, image, audio, and action commands from combinations of text, image, video, and action trajectory inputs. It serves as a foundational building block for a broad range of Physical AI applications and research spanning world understanding, world generation, simulation, and embodied policy…
Excerpt from the card by Ares Realm Studios, licensed other.
Configuration
- Architecture
- Cosmos3ForConditionalGeneration
- Context length (tokens)
- 262,144
- Layers
- 36
- Hidden size
- 4,096
- Feed-forward size
- 12,288
- Attention heads
- 32
- Key/value heads
- 8
- Head dimension
- 128
- Vocabulary size
- 151,936
- RoPE base
- 5,000,000
- Model type
- cosmos3_omni
Identity and Version
- Repository
- Ares-Realm-Studios/Cosmos3-Nano
- Publisher
- Ares Realm Studios
- Task
- Not stated by the source
- Modality
- Other
- Library
- cosmos
- Parameters
- 15.8B parameters
- Languages
- Not stated by the source
- Revision
- 6e8874bbf40f95912cd03bc40a4b595fca2aab61
- First published
- 2026-10-03
- Last updated
- 2026-10-03
Files and Weights
68 files, 35.0 GB in total. The weights are 10 files totalling 34.9 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| sound_tokenizer/diffusion_pytorch_model.safetensors | Weights | 2.0 GB | a4b6da5975bf |
| transformer/diffusion_pytorch_model-00001-of-00007.safetensors | Weights | 4.9 GB | 7d0f718198ca |
| transformer/diffusion_pytorch_model-00002-of-00007.safetensors | Weights | 5.0 GB | 7e6e7d270c19 |
| transformer/diffusion_pytorch_model-00003-of-00007.safetensors | Weights | 4.9 GB | 15d08936149a |
| transformer/diffusion_pytorch_model-00004-of-00007.safetensors | Weights | 5.0 GB | 02cebbb557bc |
| transformer/diffusion_pytorch_model-00005-of-00007.safetensors | Weights | 4.9 GB | a5e3edd742b8 |
| transformer/diffusion_pytorch_model-00006-of-00007.safetensors | Weights | 4.3 GB | 1bb99750f7ab |
| transformer/diffusion_pytorch_model-00007-of-00007.safetensors | Weights | 1.3 GB | 3f159bc8b824 |
| vae/diffusion_pytorch_model.safetensors | Weights | 1.4 GB | 230496cb59ff |
| vision_encoder/model.safetensors | Weights | 1.2 GB | b0c0f719fcf5 |
| assets/example_action_fd_agibotworld_action_chunks.json | Configuration | 48.4 KB | — |
| assets/example_action_id_av_0_output.json | Configuration | 16.0 KB | — |
| assets/example_action_id_av_1_output.json | Configuration | 15.7 KB | — |
| assets/example_i2v_prompt.json | Configuration | 9.2 KB | — |
| assets/example_reasoning_prompt.json | Configuration | 151 B | — |
| assets/example_t2v_prompt.json | Configuration | 7.7 KB | — |
| assets/example_t2vs_prompt.json | Configuration | 8.9 KB | — |
| assets/negative_prompt.json | Configuration | 15.4 KB | — |
| chat_template.json | Configuration | 5.5 KB | — |
| checkpoint.json | Configuration | 2 B | — |
| config.json | Configuration | 8.2 KB | — |
| generation_config.json | Configuration | 269 B | — |
| model.safetensors.index.json | Configuration | 113.0 KB | — |
| model_index.json | Configuration | 502 B | — |
| modular_model_index.json | Configuration | 1.7 KB | — |
| preprocessor_config.json | Configuration | 390 B | — |
| scheduler/scheduler_config.json | Configuration | 884 B | — |
| sound_tokenizer/config.json | Configuration | 1.3 KB | — |
| text_tokenizer/added_tokens.json | Configuration | 707 B | — |
| text_tokenizer/special_tokens_map.json | Configuration | 613 B | — |
| transformer/config.json | Configuration | 1.4 KB | — |
| transformer/diffusion_pytorch_model.safetensors.index.json | Configuration | 78.9 KB | — |
| vae/config.json | Configuration | 1.8 KB | — |
| video_preprocessor_config.json | Configuration | 385 B | — |
| vision_encoder/config.json | Configuration | 509 B | — |
| BIAS.md | Documentation | 4.7 KB | — |
| EXPLAINABILITY.md | Documentation | 3.2 KB | — |
| PRIVACY.md | Documentation | 1.2 KB | — |
| README.md | Documentation | 43.1 KB | — |
| SAFETY.md | Documentation | 3.7 KB | — |
| assets/example_action_fd_agibotworld_4chunk_output.mp4 | Other | 672.9 KB | 9b01266b6cd2 |
| assets/example_action_fd_agibotworld_first_frame.png | Other | 637.6 KB | 78b7288846b0 |
| assets/example_action_id_av_0_input.mp4 | Other | 1.1 MB | ff205f86ae16 |
| assets/example_action_id_av_0_output.png | Other | 135.7 KB | bc0adf920602 |
| assets/example_action_id_av_1_input.mp4 | Other | 1.6 MB | 169e65cee76e |
| assets/example_action_id_av_1_output.png | Other | 134.0 KB | 056ff91da285 |
| assets/example_i2v_input.jpg | Other | 860.8 KB | 1de51eb5c6d5 |
| assets/example_i2v_output.mp4 | Other | 22.8 MB | 35f776ae09a6 |
| assets/example_i2vs_output.mp4 | Other | 17.8 MB | 5ef623a93fe7 |
| assets/example_reasoning_input.png | Other | 230.7 KB | 6686b937bdb2 |
| assets/example_t2v_diffusers_output.mp4 | Other | 1.5 MB | b630246f9bc9 |
| assets/example_t2v_output.mp4 | Other | 9.2 MB | 8da069632812 |
| assets/example_t2v_prompt_short.txt | Other | 46 B | — |
| assets/example_t2vs_output.mp4 | Other | 4.1 MB | 3cc8ac15acd0 |
| images/benchmark-action-1.png | Other | 101.0 KB | 11772abc8515 |
| images/benchmark-overall.png | Other | 268.9 KB | 25425f6be1f1 |
| images/benchmark-reasoning.png | Other | 2.7 MB | ba95a3b243fc |
| images/benchmark-visual-audio.png | Other | 366.1 KB | 2faab808cd98 |
| text_tokenizer/chat_template.jinja | Other | 5.3 KB | — |
| .gitattributes | Repository | 2.4 KB | — |
| merges.txt | Tokenizer | 1.7 MB | — |
| text_tokenizer/merges.txt | Tokenizer | 1.7 MB | — |
| text_tokenizer/tokenizer.json | Tokenizer | 11.4 MB | aeb13307a71a |
| text_tokenizer/tokenizer_config.json | Tokenizer | 5.4 KB | — |
| text_tokenizer/vocab.json | Tokenizer | 2.8 MB | — |
| tokenizer.json | Tokenizer | 7.0 MB | — |
| tokenizer_config.json | Tokenizer | 10.9 KB | — |
| vocab.json | Tokenizer | 2.8 MB | — |
License and Download
- License
- other
- Access
- Open weights, no gate
- Download size
- 34.9 GB
Released by Ares Realm Studios through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 34.9 GB |
| 16-bit | 31.5 GB |
| 8-bit | 15.8 GB |
| 4-bit | 7.9 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Cosmos3-Nano
How much GPU memory does Cosmos3-Nano need?
About 37.8 GB at 16-bit and 9.5 GB at 4-bit: the weights (15.8B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run Cosmos3-Nano on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
What license is Cosmos3-Nano released under?
other, as its publisher declares it. Read the license text before commercial use.
What is Cosmos3-Nano's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.