Uni-Med

Closed weights Tsinghua University,Beijing University of Posts and Telecommunications 8.8B parameters November 2024

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
Tsinghua University,Beijing University of Posts and Telecommunications
Organisation type
Academia,Academia
Country
China
Published
1 November 2024
Authors
Xun Zhu, Ying Hu, Fanbin Mo, Miao Li, Ji Wu

What it does

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

Base model
Llama 2-7B,ViT-G/14

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
8.8B

7B (LLama) + 1843000000 (ViT) = 8.8B

Training data
tokens

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
1.4 × 10²³ FLOP

Finetune: 10*60*60*312000000000000*0.5=5.616e+18 Base models: 8.4e+22, 5.85e+22 Total: 1.4250562e+23

Fine-tuning compute
5.6 × 10¹⁸ 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 A800 SXM
Chips used
1
Wall-clock time
10 hours
Power draw
433 W

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
Open (restricted use)

https://github.com/MSIIP/Uni-Med (undefined license) I haven't found released pre-trained weights

How it is classified

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

Likely above 10²³ FLOP
Yes
Record confidence
Likely
Citations
23

Sources

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

Reference
Uni-Med: A Unified Medical Generalist Foundation Model For Multi-Task Learning Via Connector-MoE
Last updated
25 May 2026

What the numbers mean

Where it came from

Uni-Med was published by Tsinghua University,Beijing University of Posts and Telecommunications, in China, in November 2024. academia,Academia is the category the publisher falls under.

Its starting point was Llama 2-7B,ViT-G/14 — most models at this scale are adapted from an existing base rather than built from nothing.

This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.

Training and provenance

Training it took roughly 1.4 × 10²³ FLOP of computation, on NVIDIA A800 SXM — a measure of what producing the model cost, not of how fast it answers.

Answers

Uni-Med — common questions

01

How many parameters does Uni-Med have?

Uni-Med has 8.8B parameters. 7B (LLama) + 1843000000 (ViT) = 8.8B. 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.

02

Who created Uni-Med?

Uni-Med was published by Tsinghua University,Beijing University of Posts and Telecommunications, based in China, categorised as academia,Academia.

03

When was Uni-Med released?

Uni-Med was published in November 2024. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

04

How much compute was used to train Uni-Med?

Around 1.4 × 10²³ FLOP, on NVIDIA A800 SXM. 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.

05

What GPU do I need to run Uni-Med?

None. Uni-Med 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.

06

Is Uni-Med open source?

No. Uni-Med has not had its weights published, so it exists only as a service controlled by its owner.

Source

Original publication

Record last updated 25 May 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.