ComfyUI model file · LoRA (step distillation)

sdxl_lightning_4step_lora.safetensors

SDXL-Lightning 4-step LoRA · 4-step speed LoRA for SDXL models

This is ByteDance’s SDXL-Lightning LoRA for 4-step generation. Add it to an SDXL fine-tune and you get a full image in 4 steps instead of 25 to 40. Put it in ComfyUI/models/loras, load it with LoraLoaderModelOnly, and sample with Euler, sgm_uniform, 4 steps, CFG 1.

  • SDXL
  • 4 steps
  • 394 MB
  • ByteDance

For the plain SDXL base model, ByteDance recommends the full 4-step checkpoint instead. Use this LoRA on other SDXL models.

sdxl_lightning_4step_lora.safetensors

Type
LoRA (step distillation)
Format
Safetensors
Size
394 MB (393,854,592 bytes)
SHA256
bf56cf26…ad8506b79
ComfyUI folder
models/loras/
Released
Feb 20, 2024
SHA256 and size from Hugging Face file metadata (ByteDance/SDXL-Lightning), matching the Civitai API hash, checked by RunComfy on .

The short version

What is sdxl_lightning_4step_lora.safetensors?

SDXL-Lightning is ByteDance’s progressive adversarial distillation of SDXL. It comes in 1, 2, 4 and 8-step versions, each as a full UNet checkpoint and as a LoRA. This file is the 4-step LoRA, the most common balance of speed and quality.

Because it is a LoRA, it can turn any SDXL fine-tune (Juggernaut, RealVis, anime models) into a 4-step model. The step count must match: this LoRA at 8 steps, or the 8-step LoRA at 4, gives worse results.

Install

Download and install sdxl_lightning_4step_lora.safetensors

Where it goes
ComfyUI/
└─ models/
   ├─ …
   └─ loras/
      └─ sdxl_lightning_4step_lora.safetensors
terminal · Linux / macOS
$ cd ComfyUI/models/loras$ wget -O sdxl_lightning_4step_lora.safetensors \$   https://huggingface.co/ByteDance/SDXL-Lightning/resolve/main/sdxl_lightning_4step_lora.safetensors$ sha256sum sdxl_lightning_4step_lora.safetensors# bf56cf2657efb15e465d81402ed481d1e11c4677e4bcce1bc11fe71ad8506b79

Safetensors files hold tensors only and cannot run code. Still compare the SHA256: a failed download often leaves an HTML error page saved under the model’s filename.

Use

How to load sdxl_lightning_4step_lora.safetensors in ComfyUI

LoraLoaderModelOnly patches the base model with this LoRA. Put it between the model loader and the sampler.

Minimal graph. Click a node to open its reference page.
  • Sampler settings

    Euler sampler, sgm_uniform scheduler, 4 steps, CFG 1. Match the steps to the LoRA.

  • Strength

    Start at 1.0. Lower it a little if a fine-tune loses its style.

Run it

ComfyUI workflows that can use sdxl_lightning_4step_lora.safetensors

We match workflows by the model files they load: this exact file first, then other versions of it, then workflows built for the same use, then other models from the same family.

Built for fast sampling

1 workflow

These workflows load another fast sampling model in the same folder, or they are fast sampling workflows that already have a LoraLoaderModelOnly node. Select sdxl_lightning_4step_lora.safetensors there.

Compare

Same file, other names, and alternatives

FileRelationSizeBest for
sdxl_lightning_4step_lora.safetensorsThis page · 4-step LoRA394 MBSDXL fine-tunes, 4 steps
sdxl_lightning_8step_lora.safetensorsSame family · 8-step LoRA394 MBMore detail, 8 steps
sdxl_lightning_2step_lora.safetensorsSame family · 2-step LoRA394 MBFastest previews
sdxl_lightning_4step.safetensorsFull 4-step checkpoint6.94 GBPlain SDXL base, best quality

Troubleshoot

Common errors and fixes

Value not in list: lora_name: 'sdxl_lightning_4step_lora.safetensors' not in [...]

Why: The workflow asks for this exact filename, but ComfyUI cannot find it in models/loras. Usually the file is missing, sits in a different folder.

Fix: Move the file into ComfyUI/models/loras/, rename it to match the workflow (or pick it again in the dropdown), then refresh the node list or restart ComfyUI.

SafetensorError: Error while deserializing header

Why: The download is incomplete, or the browser saved an HTML sign-in or error page under the model’s filename.

Fix: Check that the file is exactly 393,854,592 bytes and that its SHA256 matches the value above. Download it again from Hugging Face · ByteDance/SDXL-Lightning if not.

Rather skip the setup?

Run IPAdapter V1 FaceID Plus in your browser on a cloud GPU. Nothing to download.

Try this LoRA online

FAQ

sdxl_lightning_4step_lora.safetensors FAQ

What sampler and scheduler should I use?

ByteDance’s ComfyUI instructions: Euler sampler with the sgm_uniform scheduler, 4 steps, CFG 1.

LoRA or full checkpoint?

Use the full sdxl_lightning_4step.safetensors checkpoint for the plain SDXL base model; it looks better. Use this LoRA to speed up any other SDXL model.

Where do I put sdxl_lightning_4step_lora.safetensors in ComfyUI?

In ComfyUI/models/loras/. Then refresh the node list or restart ComfyUI so LoraLoaderModelOnly lists the file.

Which ComfyUI node loads sdxl_lightning_4step_lora.safetensors?

LoraLoaderModelOnly. It is built into ComfyUI, so no custom nodes are needed.

Can I use sdxl_lightning_4step_lora.safetensors commercially?

Yes. SDXL-Lightning is released under CreativeML Open RAIL++-M, the same license as SDXL, which allows commercial use within its use restrictions.

Sources: ByteDance/SDXL-Lightning · SDXL-Lightning paper (arXiv 2402.13929) · Civitai model 350450

RunComfy
Copyright 2026 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.