# NLD-Uni

> NLD-Uni is a unified multimodal model for image understanding and image generation. It combines a diffusion language model with specialized visual components and a mixture-of-transformers design.

- **Organisation:** nvidia
- **Category:** Omni / Any-to-Any
- **Parameters:** 18.3B
- **Context length:** 131,072 tokens
- **Licence:** Unknown licence (Check licence)
- **Released:** 2026-10-05
- **Hugging Face:** 1.3K downloads (30 days), 12 likes

## Estimated memory

| Quant | ≈ GB |
|---|---|
| Q2_K | 7.44 |
| Q3_K_M | 10.29 |
| Q4_K_M | 12.39 |
| Q5_K_M | 14.27 |
| Q6_K | 16.28 |
| Q8_0 | 20.54 |
| F16 | 37.29 |

## Runs with

- [Transformers](https://github.com/huggingface/transformers)

## About

NLD-Uni is a unified multimodal model for image understanding and image generation. It combines a diffusion language model with specialized visual components and a mixture-of-transformers design. Understanding and generation use distinct transformer parameter paths within a common multimodal framework.

For image understanding, NLD-Uni encodes images into continuous visual features and generates text responses. For image generation, it models discrete visual tokens through iterative diffusion and decodes those tokens into RGB images. The architecture supports autoregressive and blockwise diffusion decoding for text, as well as image-conditioned generation and editing when enabled by the checkpoint and inference implementation.

This card describes the NLD-Uni architecture generally.

Install the dependencies in requirements.txt with Python 3.13 and CUDA 12.6. Install PyTorch before building FlashAttention:

Validated on an NVIDIA H100 80 GB. The eight BF16 weight shards contain all 1,557 active model tensors. weights_manifest.json records their hashes.

Image inputs accept Pillow images or local file paths. Image methods return Pillow images; generate returns text by default. Token IDs can be supplied through input_ids for text prompts, and return_token_ids=True returns the generated token tensor. The inference.py command exposes each task.

## Alternatives

- [google/gemma-4-12B-it](https://selfhost.directory/models/google/gemma-4-12b-it) (12B, Apache 2.0, ≈10.59 GB at Q4): Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants.
- [sensenova/SenseNova-U1.5-8B-MoT](https://selfhost.directory/models/sensenova/sensenova-u1.5-8b-mot) (17.5B, Apache 2.0, ≈12.21 GB at Q4): SenseNova-U1.5-8B-MoT is our latest native unified multimodal checkpoint for more accurate, consistent, reliable, and aesthetically compelling visual creation.
- [google/gemma-4-E4B-it](https://selfhost.directory/models/google/gemma-4-e4b-it) (8B, Apache 2.0, ≈5.9 GB at Q4): Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants.
- [google/gemma-4-12B](https://selfhost.directory/models/google/gemma-4-12b) (12B, Apache 2.0, ≈10.59 GB at Q4): Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants.
- [inclusionAI/LLaDA2.0-Uni](https://selfhost.directory/models/inclusionai/llada2.0-uni) (16.3B (1.5B active), Apache 2.0, ≈10.49 GB at Q4): LLaDA2.0-Uni: Unifying Multimodal Understanding and Generation with Diffusion Large Language Model
- [KRAFTON/A.X-K2-Raon-Speech-21B-A3B](https://selfhost.directory/models/krafton/a.x-k2-raon-speech-21b-a3b) (21.2B (3B active), CC BY-NC 4.0, ≈22.07 GB at Q4): A.X K2 Raon-Speech is a bilingual English/Korean speech language model with approximately 21.2B total parameters and 3.5B active parameters.

More: https://selfhost.directory/models/nvidia/nld-uni/alternatives

## Runs well on

- [Tesla T4](https://selfhost.directory/models/nvidia/nld-uni/on/tesla-t4): Q5_K_M, est. 14 tok/s
- [Arc A770 16GB](https://selfhost.directory/models/nvidia/nld-uni/on/arc-a770): Q5_K_M, est. 24 tok/s
- [Radeon RX 7800 XT](https://selfhost.directory/models/nvidia/nld-uni/on/rx-7800-xt): Q5_K_M, est. 26 tok/s
- [Tesla P40](https://selfhost.directory/models/nvidia/nld-uni/on/tesla-p40): Q8_0, est. 9.8 tok/s
- [Radeon RX 7900 XTX](https://selfhost.directory/models/nvidia/nld-uni/on/rx-7900-xtx): Q8_0, est. 27 tok/s
- [Instinct MI50 32GB](https://selfhost.directory/models/nvidia/nld-uni/on/instinct-mi50): Q8_0, est. 29 tok/s

## Similar size

- [ByteDance-Seed/BAGEL-7B-MoT](https://selfhost.directory/models/bytedance-seed/bagel-7b-mot) (14.7B)
- [sensenova/SenseNova-U1.5-8B-MoT](https://selfhost.directory/models/sensenova/sensenova-u1.5-8b-mot) (17.5B)
- [HiDream-ai/HiDream-E1-Full](https://selfhost.directory/models/hidream-ai/hidream-e1-full) (17.1B)
- [inclusionAI/LLaDA2.0-Uni](https://selfhost.directory/models/inclusionai/llada2.0-uni) (16.3B)
- [sensenova/SenseNova-U1.5-8B-MoT-Preview](https://selfhost.directory/models/sensenova/sensenova-u1.5-8b-mot-preview) (17.5B)
- [HiDream-ai/HiDream-E1-1](https://selfhost.directory/models/hidream-ai/hidream-e1-1) (17.1B)

## FAQ

**How much VRAM does NLD-Uni need?**
About 12 GB at Q4_K_M with an 8K context. Q8_0 needs about 21 GB and full precision about 37 GB. Longer contexts need more memory for the KV cache.

**Can NLD-Uni run on the Tesla T4?**
Yes. It fits at Q5_K_M (about 14 GB), at an estimated 14 tokens/s. It fits comfortably on the Arc A770 16GB, Radeon RX 7800 XT, Tesla P40.

**Can I use NLD-Uni commercially?**
Unclear. It is released under Unknown licence. No licence stated: assume all rights reserved. Check the licence on the model card before using it in a product.

**What is the context length of NLD-Uni?**
131,072 tokens (128K). The full window needs far more memory than an 8K context.

## Links

- huggingface: https://huggingface.co/nvidia/NLD-Uni

---
Source: https://selfhost.directory/models/nvidia/nld-uni