Image models · ByteDance
SDXL-Lightning (4-step)
- Backbone
- 2.6B
- Native output
- 1024×1024
- Default steps
- 4
- License
- openrail++
- Released
- 2024-02
How to run it
Original weights · Hugging Face
Full-precision files from ByteDance, for fine-tuning, vLLM, Transformers, Diffusers or ComfyUI.
huggingface.co/ByteDance/SDXL-Lightning (opens in a new tab)
Memory needed
Peak GPU memory at the native output size. Offloading moves the text encoder, or the whole pipeline, to system RAM between steps.
| Precision | Everything on GPU | Text encoder offloaded | Sequential offload | Download |
|---|---|---|---|---|
FP16 | 8.9 GB | 7.3 GB | 3.8 GB | Original weights (opens in a new tab) |
FP8 | 5.6 GB | 4.7 GB | 2.9 GB | Original weights (opens in a new tab) |
Components
- Backbone: 2.57B
- VAE: 0.084B
- CLIP-L: 0.123B
- OpenCLIP bigG: 0.695B
Best hardware for SDXL-Lightning (4-step)
The least expensive GPU in our catalog that runs it without offloading is the Intel Arc A380 (launch price $139) at FP8.
| Grade · points | Hardware | Precision | Memory | Time per output | Fit |
|---|---|---|---|---|---|
| 79 ptsGreat (A) · 79 of 100 points: Model quality 24.3/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 | Intel Arc B580 | FP16 | 8.9 GB / 12.0 GB | 2.9 s | Runs well |
| 79 ptsGreat (A) · 79 of 100 points: Model quality 24.3/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 | NVIDIA GeForce RTX 3060 12GB | FP16 | 8.9 GB / 12.0 GB | 2.9 s | Runs well |
| 79 ptsGreat (A) · 79 of 100 points: Model quality 24.3/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 | NVIDIA GeForce RTX 4070 SUPER | FP16 | 8.9 GB / 12.0 GB | 1.3 s | Runs well |
| 78 ptsGreat (A) · 78 of 100 points: Model quality 23.8/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 | NVIDIA GeForce RTX 4060 | FP8 | 5.6 GB / 8.0 GB | 2.3 s | Runs well |
| 77 ptsGreat (A) · 77 of 100 points: Model quality 24.3/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 12.9/15 | AMD Radeon RX 9070 XT | FP16 | 8.9 GB / 16.0 GB | 0.9 s | Runs well |
| 77 ptsGreat (A) · 77 of 100 points: Model quality 24.3/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 12.9/15 | NVIDIA GeForce RTX 4060 Ti 16GB | FP16 | 8.9 GB / 16.0 GB | 1.9 s | Runs well |
| 77 ptsGreat (A) · 77 of 100 points: Model quality 24.3/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 12.9/15 | NVIDIA GeForce RTX 4070 Ti SUPER | FP16 | 8.9 GB / 16.0 GB | 1.1 s | Runs well |
| 77 ptsGreat (A) · 77 of 100 points: Model quality 24.3/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 12.9/15 | NVIDIA GeForce RTX 4080 SUPER | FP16 | 8.9 GB / 16.0 GB | 1.0 s | Runs well |
| 77 ptsGreat (A) · 77 of 100 points: Model quality 24.3/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 12.9/15 | NVIDIA GeForce RTX 5060 Ti 16GB | FP16 | 8.9 GB / 16.0 GB | 1.8 s | Runs well |
| 77 ptsGreat (A) · 77 of 100 points: Model quality 24.3/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 12.9/15 | NVIDIA GeForce RTX 5070 Ti | FP16 | 8.9 GB / 16.0 GB | 1.1 s | Runs well |
Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.