Skip to main content
ModelRunner
bytedance avatar

SDXL Lightning 4-step API

bytedance/sdxl-lightning-4step

SDXL-Lightning is a lightning-fast text-to-image generation model that produces high-quality 1024px images in just a few steps, distilled from Stable Diffusion XL.

A100
0.00251692
GitHub
License
Weights
Paper
OpenAPI

Model Input

Input

Input prompt

Negative Input prompt

Min: 256 - Max: 1280

Width of output image. Recommended 1024 or 1280

Min: 256 - Max: 1280

Height of output image. Recommended 1024 or 1280

Min: 1 - Max: 4

Number of images to output.

scheduler

Min: 1 - Max: 10

Number of denoising steps. 4 for best results

Min: 0 - Max: 50

Scale for classifier-free guidance

Random seed. Leave blank to randomize the seed

Disable safety checker for generated images

You need to be logged in to run this model and view results.
Log in

Model Output

Output

Generated image output
Generated in 1.817 seconds
Logs (7 lines)

Model Example Requests

Examples

Example output 1Example output 2Example output 3Example output 4Example output 5

SDXL Lightning 4-step API

SDXL Lightning 4-step is a text-to-image AI model by bytedance. On ModelRunner it runs through a REST API or via MCP from any AI assistant, at about $0.00251692 per image.

POST https://queue.modelrunner.run/bytedance/sdxl-lightning-4step

cURL

# Submit a request to the queue. Input fields go at the top level of the
# body. The optional reserved "metadata" object holds your own flat string
# tags — stored on the request, never sent to the model; filter later with
# GET https://queue.modelrunner.run/requests?metadata=<url-encoded JSON>.
curl -X POST https://queue.modelrunner.run/bytedance/sdxl-lightning-4step \
  -H "Authorization: Key $MRUN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "seed": 102,
    "width": 1280,
    "height": 1024,
    "prompt": "luxury stainless-steel chronograph wristwatch on black marble, dramatic rim lighting and softbox reflections, macro 1…",
    "scheduler": "K_EULER",
    "num_outputs": 1,
    "guidance_scale": 0,
    "negative_prompt": "worst quality, low quality, noise, blurry, fingerprints, dust, scratched glass, watermark, text, logo",
    "num_inference_steps": 4,
    "disable_safety_checker": false,
    "metadata": {
      "project": "my-project"
    }
  }'
# → { "request_id": "...", "status_url": "...", "response_url": "..." }

# Poll status_url until "COMPLETED", then fetch the result
curl "https://queue.modelrunner.run/bytedance/sdxl-lightning-4step/requests/$REQUEST_ID/status" \
  -H "Authorization: Key $MRUN_API_KEY"
curl "https://queue.modelrunner.run/bytedance/sdxl-lightning-4step/requests/$REQUEST_ID" \
  -H "Authorization: Key $MRUN_API_KEY"

JavaScript

import { modelrunner } from "@modelrunner/client";

const result = await modelrunner.subscribe("bytedance/sdxl-lightning-4step", {
  input: {
    "seed": 102,
    "width": 1280,
    "height": 1024,
    "prompt": "luxury stainless-steel chronograph wristwatch on black marble, dramatic rim lighting and softbox reflections, macro 1…",
    "scheduler": "K_EULER",
    "num_outputs": 1,
    "guidance_scale": 0,
    "negative_prompt": "worst quality, low quality, noise, blurry, fingerprints, dust, scratched glass, watermark, text, logo",
    "num_inference_steps": 4,
    "disable_safety_checker": false
  },
});
console.log(result);

Python

import os
import requests

headers = {"Authorization": f"Key {os.environ['MRUN_API_KEY']}"}

submitted = requests.post(
    "https://queue.modelrunner.run/bytedance/sdxl-lightning-4step",
    headers=headers,
    json={
      "seed": 102,
      "width": 1280,
      "height": 1024,
      "prompt": "luxury stainless-steel chronograph wristwatch on black marble, dramatic rim lighting and softbox reflections, macro 1…",
      "scheduler": "K_EULER",
      "num_outputs": 1,
      "guidance_scale": 0,
      "negative_prompt": "worst quality, low quality, noise, blurry, fingerprints, dust, scratched glass, watermark, text, logo",
      "num_inference_steps": 4,
      "disable_safety_checker": false
    },
).json()

# Poll submitted["status_url"] until "COMPLETED", then:
result = requests.get(submitted["response_url"], headers=headers).json()

Input parameters

NameTypeRequiredDescription
promptstringnoInput prompt Default: "self-portrait of a woman, lightning in the background".
negative_promptstringnoNegative Input prompt Default: "worst quality, low quality".
widthintegernoWidth of output image. Recommended 1024 or 1280 Default: 1024.
heightintegernoHeight of output image. Recommended 1024 or 1280 Default: 1024.
num_outputsintegernoNumber of images to output. Default: 1.
schedulerenumnoscheduler Default: "K_EULER".
num_inference_stepsintegernoNumber of denoising steps. 4 for best results Default: 4.
guidance_scalenumbernoScale for classifier-free guidance Default: 0.
seedintegernoRandom seed. Leave blank to randomize the seed Default: 0.
disable_safety_checkerbooleannoDisable safety checker for generated images Default: false.

Machine-readable: OpenAPI schema · llms.txt

Use SDXL Lightning 4-step from Claude & Cursor (MCP)

Point Claude Code, Claude Desktop, Cursor, or any MCP client at the ModelRunner MCP server and SDXL Lightning 4-step becomes a tool your assistant can call directly — it authorizes via OAuth (no API key in config) and runs this model with the run_model tool using the endpoint bytedance/sdxl-lightning-4step.

MCP client config (Claude Desktop, Cursor)

{
  "mcpServers": {
    "modelrunner": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.modelrunner.run/mcp"]
    }
  }
}

Claude Code

claude mcp add --transport http modelrunner https://mcp.modelrunner.run/mcp

Then ask your assistant, for example: “Run bytedance/sdxl-lightning-4step on ModelRunner to generate image”. MCP setup guide.

Model Details

Model Details

# SDXL-Lightning

**SDXL-Lightning** is a state-of-the-art text-to-image generation model designed for both **speed** and **quality**, capable of producing **high-resolution 1024px images** in remarkably few steps.

Built upon the foundation of [`stabilityai/stable-diffusion-xl-base-1.0`](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0), it leverages *Progressive Adversarial Diffusion Distillation* to achieve rapid yet powerful image synthesis.

## Features - 🚀 **Lightning-fast inference** with 1-step, 2-step, 4-step, and 8-step distilled models. - **2-step, 4-step, and 8-step** models deliver outstanding image quality. - **1-step** model is experimental, exploring the limits of extreme efficiency.

- 🧩 **Model formats**: - **Full UNet checkpoints** – deliver the best image quality. - **LoRA checkpoints** – lightweight and easily applied to other base models.

## Open Source We release SDXL-Lightning fully open-sourced as part of our research to empower the community with cutting-edge fast diffusion technology. For more details, please see our paper: **SDXL-Lightning: Progressive Adversarial Diffusion Distillation**.