Cosmos3-Edge

Checkpointnvidia/Cosmos3-EdgeModel card indexedModel card

Support summary

Cosmos3-Edge is a checkpoint published by NVIDIA, built on the Cosmos3EdgeForConditionalGeneration architecture. vLLM 0.28.0, the newest indexed vLLM release, documents support.

vLLM documents support in 4 of 20 indexed releases, from 0.26.0 to 0.28.0.

Configured context
131,072 tokens
License
openmdw1.1-license
Hugging Face model card
Publisher
NVIDIA
Library
cosmos
Architecture
Cosmos3EdgeForConditionalGeneration
Weight tensor elements
3.9B
Tensor types
BF16
Configured context
131,072 tokens
Weight access
Public
Repository created
2026-07-01
Source 1: nvidia/Cosmos3-Edge · publisher model cardSource 2: nvidia/Cosmos3-Edge · Hugging Face metadataSource 3: nvidia/Cosmos3-Edge · publisher configurationReviewed
Model notes and sources

Engine support

3 engines

Engine support for Cosmos3-Edge
EngineReleaseCheckpoint supportEvidence scopeLast known supported
vLLM0.28.0Newest indexedCheckpoint evidenceSource: vLLM v0.28.0 · docs/models/supported_models.mdReviewed 0.28.0This release
Features, dependencies and evidence for vLLM 0.28.0
SGLang0.5.19Newest indexedNo matching evidence curatedNo supported release indexed
Features, dependencies and evidence for SGLang 0.5.19
TensorRT-LLM1.2.1Newest indexedNo matching evidence curatedNo supported release indexed
Features, dependencies and evidence for TensorRT-LLM 1.2.1

Recipes and cookbooks