---
title: InternLumina-U2
canonical_url: "https://www.modelscope.cn/models/Shanghai_AI_Laboratory/InternLumina-U2"
md_url: "https://www.modelscope.cn/models/Shanghai_AI_Laboratory/InternLumina-U2.md"
repository: Shanghai_AI_Laboratory/InternLumina-U2
last_updated: 2026-09-25
license: apache-2.0
parameters: 18.8B
tensor_type:
  - BF16
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
downloads: 50
stars: 2
tags:
  - diffusion
  - multimodal
  - unified-model
  - image-generation
  - image-editing
  - vision-language
---

# InternLumina-U2

> InternLumina-U2 - Shanghai_AI_Laboratory 在 ModelScope 开源的模型。A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding, Image Generation and Editing.

Shanghai_AI_Laboratory/InternLumina-U2 是 ModelScope 魔搭社区上的 18.8B 参数机器学习模型，采用 apache-2.0 许可。

- **Repository**: Shanghai_AI_Laboratory/InternLumina-U2
- **License**: apache-2.0
- **Parameters**: 18.8B
- **Tags**: diffusion, multimodal, unified-model, image-generation, image-editing, vision-language
- **Downloads**: 50
- **Stars**: 2
- **Last updated**: 2026-09-25

Source: https://www.modelscope.cn/models/Shanghai_AI_Laboratory/InternLumina-U2

---

# Intern Lumina U2

A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding, Image Generation and Editing.

[📑 Technical Report (Coming Soon)] · [🌐 Project Page](https://internlm.github.io/InternLumina-U2/) · [💻 Code](https://github.com/InternLM/InternLumina-U2)

## Overview

Intern Lumina U2 is a unified multimodal model that brings language, image, video and 3D into a
single framework, covering text QA, text-to-image generation, image understanding, image editing,
video understanding and 3D understanding with one model. It is a 16B-parameter MoE with 1B active
parameters (16B-A1B), pairing an efficient sparse backbone with an 8-codebook fully-discrete visual
representation built on [AToken](https://github.com/apple/ml-atoken).

## Checkpoints in this repo

This repo hosts checkpoints of the same architecture trained on different hardware stacks:

- `ascend/` — trained on Huawei Ascend NPUs.
- `nvidia/` — trained on NVIDIA GPUs.

## Usage

Inference code, setup instructions and examples live in the GitHub repo:
https://github.com/InternLM/InternLumina-U2. Point `CHECKPOINT` at one of the subfolders above
when running `infer_1024_sft.sh`.

## Benchmarks

Preliminary, partial results. Full comparison tables will appear in the upcoming technical report.
See the [GitHub repo](https://github.com/InternLM/InternLumina-U2#-benchmarks) for the current table.

## License

Apache 2.0.

## Citation

```bibtex
@misc{internluminau2,
  title  = {Intern Lumina U2},
  author = {{Intern Lumina U2 Team, Shanghai AI Laboratory}},
  year   = {2026},
  note   = {Tech report coming soon}
}
```