Open-weight model
cosmos-libero-noood-baseline-gbs2048-4k
by Cosmos-Data Cosmos-Data/cosmos-libero-noood-baseline-gbs2048-4k
cosmos-libero-noood-baseline-gbs2048-4k is an open-weight model from Cosmos-Data. It has 15.2B parameters. At 16-bit it needs about 36.4 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index.
Research artifacts. Details will be provided after publication.
Runs On
What it takes to serve cosmos-libero-noood-baseline-gbs2048-4k (15.2B 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 | 30.3 GB | 36.4 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 15.2 GB | 18.2 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 7.6 GB | 9.1 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 20, 2026.
Model Card
Research artifacts. Details will be provided after publication.
Excerpt from the card by Cosmos-Data.
Configuration
- Architecture
- Cosmos3OmniModel
- Model type
- cosmos3_omni
Identity and Version
- Repository
- Cosmos-Data/cosmos-libero-noood-baseline-gbs2048-4k
- Publisher
- Cosmos-Data
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- 15.2B parameters
- Languages
- Not stated by the source
- Revision
- 3969a532d0856f58d18fa8ca230e342dd18b9e1f
- First published
- 2026-10-12
- Last updated
- 2026-10-13
Files and Weights
25 files, 121.4 GB in total. The weights are 7 files totalling 30.3 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00007.safetensors | Weights | 4.9 GB | 702f054ece99 |
| model-00002-of-00007.safetensors | Weights | 5.0 GB | fb060d8dcd79 |
| model-00003-of-00007.safetensors | Weights | 4.9 GB | 4c4155fe2dba |
| model-00004-of-00007.safetensors | Weights | 5.0 GB | ed3a0e22c98a |
| model-00005-of-00007.safetensors | Weights | 4.9 GB | 0f13d9bec749 |
| model-00006-of-00007.safetensors | Weights | 4.3 GB | e286c034b2e6 |
| model-00007-of-00007.safetensors | Weights | 1.3 GB | 1680964c0978 |
| chat_template.json | Configuration | 5.5 KB | — |
| checkpoint.json | Configuration | 3 B | — |
| config.json | Configuration | 6.3 KB | — |
| model.safetensors.index.json | Configuration | 89.6 KB | — |
| preprocessor_config.json | Configuration | 390 B | — |
| training_checkpoint/config.yaml | Configuration | 21.1 KB | — |
| video_preprocessor_config.json | Configuration | 385 B | — |
| README.md | Documentation | 64 B | — |
| training_checkpoint/checkpoints/iter_000004000/model/__0_0.distcp | Other | 22.8 GB | 60e56c4b7179 |
| training_checkpoint/checkpoints/iter_000004000/model/__1_0.distcp | Other | 22.8 GB | 9d447b8dacf1 |
| training_checkpoint/checkpoints/iter_000004000/model/__2_0.distcp | Other | 22.8 GB | bcbbdb9f5712 |
| training_checkpoint/checkpoints/iter_000004000/model/__3_0.distcp | Other | 22.8 GB | 183ff8c61f5b |
| .gitattributes | Repository | 2.0 KB | — |
| training_checkpoint/checkpoints/iter_000004000/model/.metadata | Repository | 1.5 MB | 77d7e9b2c8b6 |
| merges.txt | Tokenizer | 1.7 MB | — |
| tokenizer.json | Tokenizer | 7.0 MB | — |
| tokenizer_config.json | Tokenizer | 10.9 KB | — |
| vocab.json | Tokenizer | 2.8 MB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 30.3 GB
Released by Cosmos-Data through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 30.3 GB |
| 16-bit | 30.3 GB |
| 8-bit | 15.2 GB |
| 4-bit | 7.6 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About cosmos-libero-noood-baseline-gbs2048-4k
How much GPU memory does cosmos-libero-noood-baseline-gbs2048-4k need?
About 36.4 GB at 16-bit and 9.1 GB at 4-bit: the weights (15.2B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run cosmos-libero-noood-baseline-gbs2048-4k 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.