⚡ Quick Answer
A .safetensors file is a format for storing AI model weights — checkpoints, LoRAs, VAEs, and more. It loads faster than older formats and, unlike .ckpt files, can't execute hidden code when opened.
Before safetensors existed, most Stable Diffusion models shipped as .ckpt files built on Python's pickle format — a format that can embed and run arbitrary code the moment it's loaded. Hugging Face built safetensors specifically to store only tensor data, with no code execution path at all, while also loading noticeably faster.
Where You'll See It
Almost every model folder in ComfyUI: models/checkpoints, models/loras, models/vae, models/clip, and models/controlnet all expect .safetensors files by default, and any "Load" node in a workflow — Load Checkpoint, Load LoRA, Load VAE — will show a dropdown of the .safetensors files sitting in that folder.
Quick Example
flux1-dev.safetensors is the main Flux.1 Dev checkpoint file, and it's a safetensors file like nearly everything else you'll download from Hugging Face or Civitai today.
Common Mix-Up
Frequently Asked Questions
See It In Action
Ready to install your first models?
Our install guide walks through downloading and placing checkpoints, LoRAs, and VAEs correctly.
Published: 2026-09-04 · Last updated: 2026-09-04
Join the discussion
Sign in to leave a comment or reply
No comments yet
Be the first to share your thoughts!
