SDXL Lightning needs about 3.8 GB of memory at Q5_0 (a 3.0 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 15 quants, from 2.6 GB (Q4_0) to 4.1 GB (Q8_0).
SDXL Lightning – a compact 3.47B parameter diffusion model
SDXL Lightning is a 3.47 billion parameter diffusion model available in multiple GGUF quantizations. It is designed for image generation tasks and can be run with llama.cpp. The model offers a range of quant options to fit different memory budgets.
Good at
- Fast inference with low‑memory quants
- High‑quality image generation
- Versatile quant support
Limits
- No vision support in this text‑only spec
- Requires 8k context for memory estimates
- License not specified
Best for
- Generating images from prompts
- Fine‑tuning on custom datasets
- Running on edge GPUs
Who it’s for. Developers and researchers who need a lightweight diffusion model for image generation. The model is suitable for local deployment with llama.cpp and various quantizations.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_0 | 2.6 GB | 3.4 GB | Fits | Get file → | |
| Q4_0 | 2.6 GB | 3.4 GB | Fits | Get file → | |
| Q4_0 | 2.6 GB | 3.4 GB | Fits | Get file → | |
| Q4_1 | 2.8 GB | 3.6 GB | Fits | Get file → | |
| Q4_1 | 2.8 GB | 3.6 GB | Fits | Get file → | |
| Q4_1 | 2.8 GB | 3.6 GB | Fits | Get file → | |
| Q5_0 | 3.0 GB | 3.8 GB | Fits | Get file → | |
| Q5_0 | 3.0 GB | 3.8 GB | Fits | Get file → | |
| Q5_0 | 3.0 GB | 3.8 GB | Fits | Get file → | |
| Q5_1 | 3.2 GB | 4.0 GB | Fits | Get file → | |
| Q5_1 | 3.2 GB | 4.0 GB | Fits | Get file → | |
| Q5_1 | 3.2 GB | 4.0 GB | Fits | Get file → | |
| Q8_0 | 4.1 GB | 5.1 GB | Fits | Get file → | |
| Q8_0 | 4.1 GB | 5.1 GB | Fits | Get file → | |
| Q8_0 | 4.1 GB | 5.1 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_0 (file size 2.58 GB, memory 3.4 GB). For 12 GB VRAM, Q4_1 (2.77 GB, 3.6 GB) or Q5_0 (2.96 GB, 3.8 GB). For 16 GB VRAM, Q5_1 (3.15 GB, 4 GB) or Q8_0 (4.1 GB, 5.1 GB). Memory estimates are approximate.
What GPU do you need for SDXL Lightning?
These figures add the file size, an fp16 KV cache at an 8K context window and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is Q5_1 (3.2 GB file, about 4.0 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (4.1 GB file, about 5.1 GB in use).
Which SDXL Lightning quant should you choose?
Q5_0 (3.0 GB) is the middle option in this repository. Closer to the original model than 4-bit, for a moderate increase in size. A good pick when you have memory to spare.
If memory is tight, the smallest file is Q4_0 at 2.6 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about SDXL Lightning
How much VRAM does SDXL Lightning need?
About 3.8 GB at Q5_0 with an 8K context window. That is the 3.0 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run SDXL Lightning on an 8 GB GPU?
Yes, with Q8_0: the 4.1 GB file needs about 5.1 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of SDXL Lightning GGUF?
It depends on the quant: Q4_0 is 2.6 GB, Q4_0 is 2.6 GB, Q4_0 is 2.6 GB, Q4_1 is 2.8 GB, Q4_1 is 2.8 GB, Q4_1 is 2.8 GB and 9 more.
Where can I download SDXL Lightning GGUF?
The files are on Hugging Face in the mzwing/SDXL-Lightning-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
Can I run SDXL Lightning on 8 GB VRAM?
Yes, the Q4_0 quant uses 3.4 GB memory at 8k context, fitting within 8 GB VRAM.
What quant gives the best image quality?
Higher quants such as Q8_0 provide the most accurate results but require more memory.
Is there a license restriction?
The license is not specified in the provided data.