UniDiffuser (多模态大模型)
No estimate
No hardware requirements for this model
This model's weights are open, but no parameter count has been published for it. Every memory and speed figure starts from that number, so we would rather show nothing than a fabricated estimate.
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
- ShengShu,Tsinghua University,Beijing Academy of Artificial Intelligence / BAAI
- Organisation type
- Industry,Academia,Academia
- Country
- China
- Published
- 30 May 2023
- Authors
- Fan Bao, Shen Nie, Kaiwen Xue, Chongxuan Li, Shi Pu, Yaole Wang, Gang Yue, Yue Cao, Hang Su, Jun Zhu
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Image generation
- Task
- Text-to-image, Image captioning, Image generation, Image-to-image
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.
- Training data
- tokens
"The training is multiple-staged following Stable Diffusion (Rombach et al., 2022). In the first stage, we train 250K steps at 256×256 resolution on laion2B-en with a batch size of 11264 and 5K warm-up steps. In the second stage, we fine-tune the model with 200K steps at 512×512 resolution on laion-high-resolution with a batch size of 2112 and 5K warm-up steps. In the last stage, we resume from the last checkpoint of the second stage (including both weights of the model and states of the optimiz…
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
- 2 × 10²² FLOP
- How it was established
- Hardware
312000000000000 FLOP/GPU/sec * 672 hours * 3600 sec / hour * 88 GPUs * 0.3 [assumed utilization] = 1.992646656e+22 FLOP
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 SXM4 80 GB
- Chips used
- 88
- Wall-clock time
- 672 hours (28 days)
- Power draw
- 70.1 kW
"The training takes around 28 days on 88 A100 (80GB) GPUs" 28 days = 672 hours
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
- Open — downloadable
- Model access
- Open weights (unrestricted)
- Training code
- Unreleased
- Hugging Face
- thu-ml
AGPL-3.0 https://github.com/thu-ml/unidiffuser https://huggingface.co/thu-ml/unidiffuser-v1
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Record confidence
- Confident
Sources
Where this record came from and when it was last checked.
- Reference
- One Transformer Fits All Distributions in Multi-Modal Diffusion at Scale
- Last updated
- 11 February 2026
What the numbers mean
Where it came from
UniDiffuser (多模态大模型) was published by ShengShu,Tsinghua University,Beijing Academy of Artificial Intelligence / BAAI, in China, in May 2023. industry,Academia,Academia is the category the publisher falls under.
It works in Image generation, and is recorded as doing text-to-image, Image captioning, Image generation, Image-to-image.
The weights being open is what puts this page in the calculator rather than only in the catalogue: it is a model you can actually hold. It is published under the thu-ml organisation on Hugging Face.
Training and provenance
The training run consumed about 2 × 10²² FLOP, on NVIDIA A100 SXM4 80 GB. That figure describes the cost of creating it and has no bearing on how quickly it generates text.
Answers
UniDiffuser (多模态大模型) — common questions
What GPU do I need to run UniDiffuser (多模态大模型)?
We cannot say. UniDiffuser (多模态大模型) has open weights, but no parameter count has been published for it, and every memory and speed calculation starts from that number. We would rather show nothing than a fabricated estimate.
Is UniDiffuser (多模态大模型) open source?
Its weights are published, so UniDiffuser (多模态大模型) can be downloaded and run on your own hardware. Note that open weights is not the same as open source in the full sense — it says nothing about the training data, the training code, or the commercial terms attached.
How many parameters does UniDiffuser (多模态大模型) have?
No parameter count has been published for UniDiffuser (多模态大模型), which is why no memory or speed figure appears on this page.
Who created UniDiffuser (多模态大模型)?
UniDiffuser (多模态大模型) was published by ShengShu,Tsinghua University,Beijing Academy of Artificial Intelligence / BAAI, based in China, categorised as industry,Academia,Academia.
When was UniDiffuser (多模态大模型) released?
UniDiffuser (多模态大模型) was published in May 2023. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.
What is UniDiffuser (多模态大模型) used for?
UniDiffuser (多模态大模型) works in Image generation, and is recorded as handling text-to-image, Image captioning, Image generation, Image-to-image. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.
Where can I download UniDiffuser (多模态大模型)?
Its weights are published under the thu-ml organisation on Hugging Face. We do not host model files — this site calculates what hardware is needed to run them.
How much compute was used to train UniDiffuser (多模态大模型)?
Around 2 × 10²² FLOP, on NVIDIA A100 SXM4 80 GB. 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.
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.