NÜWA

Closed weights Microsoft Research,Peking University 870M parameters November 2021

No estimate

No hardware requirements for this model

The weights for this model have not been published, so it cannot be downloaded or run on your own hardware at any size. It is reachable only through its provider, and no graphics card changes that.

On record

Full specification

Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.

Origin

Who built this model, where, and when it was published.

Organisation
Microsoft Research,Peking University
Organisation type
Industry,Academia
Country
United States of America, China
Published
24 November 2021
Authors
Chenfei Wu, Jian Liang, Lei Ji, Fan Yang, Yuejian Fang, Daxin Jiang, Nan Duan

What it does

The problem areas the model was built for. A model can carry several of each.

Domain
Multimodal, Vision, Image generation, Video, Language
Task
Image generation, Video generation, Text-to-image, Text-to-video
Approach
Self-supervised learning

Size

How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.

Parameters
870M

Section 4.1

Training data
5,547,780,000 tokens

we first pre-train N ̈UWA on three datasets: Conceptual Captions [22] for text-to-image (T2I) generation, which includes 2.9M text-image pairs, Mo- ments in Time [26] for video prediction (V2V), which in- cludes 727K videos, and VATEX dataset [43] for text-to- video (T2V) generation, which includes 241K text-video pairs.

Training compute

The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.

Training compute
7.2 × 10²¹ FLOP

From AI Tracker: "Compute cost: End of Sec 4.1: "We pre-train on 64 A100 GPUs for two weeks". Half precision FLOPs of A100: 312000000000000 64 gpus *312000000000000 FLOPs *0.3 utilization * 14 day* (24*60*60) seconds / day=7.245988e+21

How it was established
Hardware

The training run

What it physically took to train: which chips, how many, for how long, and what that drew from the wall.

Training hardware
NVIDIA A100

Availability

Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.

Weights
Closed — provider access only
Model access
Unreleased
Training code
Unreleased

https://github.com/microsoft/NUWA

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Why it is tracked
SOTA improvement

"NÜWA achieves state-of-the-art results on text-to-image generation, text-to-video generation, video prediction, etc"

Record confidence
Confident
Citations
340

Sources

Where this record came from and when it was last checked.

Reference
NÜWA: Visual Synthesis Pre-training for Neural visUal World creAtion
Last updated
1 January 2026

What the numbers mean

Where it came from

NÜWA was published by Microsoft Research,Peking University, in the country recorded as United States of America, during November 2021. The category the publisher falls under is industry,Academia.

It works in the domain of Multimodal, Vision, Image generation, Video, Language, and is recorded as performing the task of image generation, Video generation, Text-to-image, Text-to-video.

Its weights were never published, so it can only be reached through its provider. No graphics card changes that.

Training and provenance

Producing it required arithmetic totalling around 7.2 × 10²¹ FLOP, on hardware recorded as NVIDIA A100. That figure measures what producing the model cost, and has no bearing on how fast it answers.

It was trained on a corpus of about 5,547,780,000 tokens of text.

Its inclusion criterion: sOTA improvement.

Answers

NÜWA — common questions

01

NÜWA— is it open source?

No. Its weights have not been published, so it exists only as a service controlled by its owner.

02

NÜWA— how many parameters does it have?

It has a parameter count of 870M. Section 4.1. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.

03

NÜWA— who created it?

It was published by Microsoft Research,Peking University, based in United States of America, an organisation categorised as industry,Academia.

04

NÜWA— when was it released?

It was published in November 2021. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

05

NÜWA— what is it used for?

It works in the domain of Multimodal, Vision, Image generation, Video, Language, and is recorded as handling the task of image generation, Video generation, Text-to-image, Text-to-video. These are the areas it was designed around; they describe intent rather than a hard boundary.

06

NÜWA— how much compute was used to train it?

Training consumed around 7.2 × 10²¹ FLOP, on hardware recorded as NVIDIA A100. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.

07

NÜWA— what GPU do I need to run it?

None. This is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.

Source

Original publication

Record last updated 1 January 2026

The other direction

Looking at it from the other side?

This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.