SDXL-Lightning 4-step LoRA: identify the adapter, not the full checkpoint
Identify the official 4-step adapter by base family and file. Keep it separate from the repository’s full checkpoints and other step variants.
Version and base-model identity
Author: ByteDance · Acceleration / distillation
sdxl-lightning-4step
SDXL · author-declared
Trigger words: Not recorded
Author step variant: 4
sdxl-lightning-4step ↗Publisher metadata; weights not downloaded or executed. Check the version-specific workflow and license.
License source ↗File and download boundary: ComfyUI World provides source references and metadata only. Model and LoRA weights, node packages, and workflow files stay with their original providers; this site does not store, mirror, or proxy them. Downloads and access conditions are controlled by each provider.
Identify the exact resource
This entry covers ByteDance’s sdxl_lightning_4step_lora.safetensors, not the similarly named full model or UNet. The author publishes those forms separately and places the ComfyUI adapter in models/loras. R01 The version record preserves the publisher’s file checksum; this site has not downloaded and rehashed the weights. R02
Match the base before tuning settings
The catalog records the SDXL base for this version. It is not a Qwen Image 2.1 adapter. Similar filenames, identical extensions and a shared image-generation task do not establish cross-architecture compatibility. Identify the checkpoint you actually load and follow the workflow for this adapter version. The author card has conflicting wording: its LoRA prose says “non-SDXL base models,” while its LoRA example loads Stable Diffusion XL base 1.0 and the same card says the models were distilled from SDXL. Do not take that phrase as proof of cross-architecture compatibility. The author also recommends the full checkpoint for an SDXL-base path when quality is the priority; use the format and settings matching your chosen workflow. R01
Keep a useful comparison
Preserve a working graph without the adapter. Add this single adapter before combining accelerators or changing multiple controls. Fix the input and record loader messages and outputs. Change other settings one at a time. This is an editorial comparison procedure, not a performance benchmark.
The author pairs weights with their step counts and describes an Euler / sgm_uniform ComfyUI path. R01 The four-step label identifies the variant; it is not a measured speedup across all hardware.
Record unknowns honestly
An empty trigger-word list means that a specific trigger was not recorded in this review. It is not a universal statement about derived versions. Unrecorded strength, VRAM and success-rate fields are likewise unknown, not zero. Check file identity and licensing before using the resource, and retain the actual base, settings and workflow when sharing a result.
Sources and verification
Original source and code review: 2026-09-22; publisher card, pinned file SHA256 and license rechecked 2026-09-25. No GPU run, weight download or quality benchmark. Procedures are editorial guidance, not a confirmed diagnosis of your environment.
Related reading
Documented dependencies
This page does not inspect your machine or guarantee execution.
File names describe recorded requirements. A linked page may be a tutorial or repository, not a direct download; verify the exact file, revision, access terms and checksum at the publisher.
Was this page helpful?
Anonymous. Only a yes/no counter is stored; no account, IP address or device details.
Sources & references
Source and code review on 2026-09-22. No GPU run, weight download or quality benchmark. Procedures are editorial guidance, not a confirmed diagnosis of your environment.
01ByteDance SDXL-Lightning author model cardSource checked: 2026-09-2502SDXL-Lightning 4-step LoRA fileSource checked: 2026-09-2503ByteDance SDXL-Lightning licenseSource checked: 2026-09-25