⚡ Quick Answer
A checkpoint is the main AI model file in ComfyUI. It's the "brain" that generates your image — it decides what styles, subjects, and details the model can produce. Every ComfyUI workflow starts by loading one.
A checkpoint (sometimes called a model file) is a large file — usually 2GB to 7GB — that contains everything an AI model learned during training. Swap the checkpoint and the same prompt produces completely different results: one checkpoint might generate realistic photos, another anime art, another 3D renders.
Where You'll See It
The Load Checkpointnode is almost always the first node in a ComfyUI workflow. Drag a checkpoint file into ComfyUI, or select it from the node's dropdown, and it outputs three connections:
- MODEL — feeds into your sampler, this is the actual image-generation engine
- CLIP — feeds into your prompt/text encoder nodes, this is how the model reads your prompt
- VAE — feeds into your image decoder, this turns the AI's internal data into a viewable image
Quick Example
sd_xl_base_1.0.safetensors is a checkpoint for SDXL. flux1-dev.safetensors is a checkpoint for FLUX. Both go in your ComfyUI/models/checkpoints/ folder and load the same way, but they produce very different results because they were trained differently.
Checkpoint vs LoRA — the Common Mix-Up
A checkpoint is the whole model. A LoRA is a small add-on file (usually under 200MB) that nudges an existing checkpoint toward a specific style or subject, without replacing it. You always need a checkpoint loaded first — a LoRA has nothing to attach to without one.
Frequently Asked Questions
See It In Action
Ready to pick your first checkpoint?
Our full guide compares the most popular checkpoints and which fits your hardware.
Published: 2026-08-25 · Last updated: 2026-08-25
Join the discussion
Sign in to leave a comment or reply
No comments yet
Be the first to share your thoughts!