Video models · NVIDIA

Cosmos-Predict2 2B Video2World

Name: Cosmos-Predict2 2B Video2World
Author: NVIDIA

Backbone: 2.0B
Native output: 1280×704, 93 frames
Default steps: 35
License: nvidia-open-model-license
Released: 2025-06

How to run it

Original weights · Hugging Face

Full-precision files from NVIDIA, for fine-tuning, vLLM, Transformers, Diffusers or ComfyUI.

huggingface.co/nvidia/Cosmos-Predict2-2B-Video2World (opens in a new tab)

Free Hugging Face account needed: accept the license on the model page and the download starts right away.

Official repo · GitHub

Reference code, inference scripts and docs from NVIDIA. The weights themselves download from Hugging Face.

github.com/nvidia-cosmos/cosmos-predict2 (opens in a new tab)

git clone https://github.com/nvidia-cosmos/cosmos-predict2

Links checked 1 day ago against Hugging Face, Ollama and GitHub.

Apps that run Cosmos-Predict2 2B Video2World

Our pick for each kind of computer, assuming the model fits in its memory. Pick your own GPU or device in the Model Finder for an exact answer.

Mac (Apple silicon)

ComfyUI

Runs this model with built-in or popular custom nodes; GGUF files load with ComfyUI-GGUF.

Also good:

Windows PC, NVIDIA GPU

ComfyUI

Runs this model with built-in or popular custom nodes; GGUF files load with ComfyUI-GGUF.

Also good:

Linux, NVIDIA or AMD GPU

ComfyUI

Runs this model with built-in or popular custom nodes; GGUF files load with ComfyUI-GGUF.

Also good:

App	Type	Opens it as	Systems	Engine	Start
ComfyUI	Desktop app	Built-in support	macOSWindowsLinux	PyTorch, GGUF via ComfyUI-GGUF	Get ComfyUI
SwarmUI	Desktop app	Built-in support	macOSWindowsLinux	ComfyUI	Get SwarmUI
SD.Next	Desktop app	Built-in support	macOSWindowsLinux	PyTorch	Get SD.Next
Stability Matrix	Desktop app	Built-in support	macOSWindowsLinux	installs ComfyUI, Forge, InvokeAI, SwarmUI, SD.Next or Fooocus	Get Stability Matrix

Which apps can open a model depends on the files published for it: GGUF for llama.cpp-based apps, MLX for Apple’s engine, the original weights for vLLM and SGLang. Compare every app.

Memory needed

Peak GPU memory at the native output size. Offloading moves the text encoder, or the whole pipeline, to system RAM between steps.

Precision	Everything on GPU	Text encoder offloaded	Sequential offload	Download
`FP16`	33.6 GB	24.0 GB	21.2 GB	Original weights (opens in a new tab)
`FP8`	26.8 GB	22.0 GB	20.5 GB	Original weights (opens in a new tab)

Components

Backbone: 2B
VAE: 0.127B
T5-XXL: 4.76B

Best hardware for Cosmos-Predict2 2B Video2World

The least expensive GPU in our catalog that runs it without offloading is the Intel Arc Pro B70 (launch price $949) at FP8.

Quality	Hardware	Quant	Memory	Speed	Runs
56Good quality (B) · 56 of 100: how capable the model is at FP8, whatever it runs on. Match score 80/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GeForce RTX 5090	`FP8`	26.8 GB / 32.0 GB	3.1 min	Runs great
58Good quality (B) · 58 of 100: how capable the model is at FP16, whatever it runs on. Match score 74/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA RTX PRO 6000 Blackwell	`FP16`	33.6 GB / 96.0 GB	3.1 min	Runs great
56Good quality (B) · 56 of 100: how capable the model is at FP8, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GeForce RTX 4090	`FP8`	22.0 GB / 24.0 GB	3.9 min	Runs welltight fit · encoder in RAM
56Good quality (B) · 56 of 100: how capable the model is at FP8, whatever it runs on. Match score 70/100 (quality, fit, speed and memory use) sets the ranking.	AMD Radeon RX 7900 XTX	`FP8`	22.0 GB / 24.0 GB	4.4 min	Runs welltight fit · encoder in RAM
58Good quality (B) · 58 of 100: how capable the model is at FP16, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking.	AMD Ryzen AI Max+ 395 (128 GB)	`FP16`	33.6 GB / 96.0 GB	8.0 min	Runs well
56Good quality (B) · 56 of 100: how capable the model is at FP8, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GeForce RTX 3090	`FP8`	22.0 GB / 24.0 GB	7.4 min	Runs welltight fit · encoder in RAM
58Good quality (B) · 58 of 100: how capable the model is at FP16, whatever it runs on. Match score 67/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA RTX Spark N1X (6144-core GPU) (128 GB)	`FP16`	33.6 GB / 102 GB	9.2 min	Runs well
58Good quality (B) · 58 of 100: how capable the model is at FP16, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GB10 (128 GB)	`FP16`	33.6 GB / 115 GB	8.7 min	Runs well
56Good quality (B) · 56 of 100: how capable the model is at FP8, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA A100 40GB	`FP8`	26.8 GB / 40.0 GB	3.9 min	Runs great
56Good quality (B) · 56 of 100: how capable the model is at FP8, whatever it runs on. Match score 77/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA RTX 5000 Ada Generation	`FP8`	26.8 GB / 32.0 GB	4.5 min	Runs well

Check it on your own hardware

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.