nvidia / text-to-image updated 3 weeks ago

Cosmos3-Super-Text2Image-4Step

Name: Cosmos3-Super-Text2Image-4Step
Author: nvidia

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.

Params

64.0 B

Context

—

Downloads 30d

102 K

Likes

Licence: other Not gated SAFETENSORS View on Hugging Face ↗

Download history

daily snapshots · 16 days

102 K21 K

Aug 10Aug 15Aug 20Aug 25

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

File	Quant	Size	Est. VRAM	Verdict on RTX 4090 · 24 GB
model.safetensors	bf16	131.4 GB	154.6 GB	❌ Won’t fit

Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Specifications

Architecture: Cosmos3ForConditionalGeneration
Parameters: 64.0 B
Tensor type: BF16
Licence: other
First seen on the Hub: 2026-07-07
Training datasets: undisclosed
Added to our catalog: 2026-08-10

Compare with

Flux2-Klein-9B-True-V2 696 K Qwen-Image-Lightning 441 K Pony_Diffusion_V6_XL 274 K Z-Image-Lora 236 K Juggernaut-XL-v9 234 K IP-Adapter-FaceID 215 K

Appears in

Best open coding models Commercial-use text models Models that fit in 12 GB