Where Do Safetensors Go in ComfyUI? Folder Map
Find where to put .safetensors files in ComfyUI: checkpoints, LoRA, VAE, ControlNet, Flux diffusion models, text encoders, and upscalers.
Start with Wonderful Launcher if this issue touches your real ComfyUI environment. Use the docs to understand the fix, and use the app to inspect the machine you already have.
Download Wonderful LauncherTested Environment
- OS: Windows 10 / 11
- Launcher: Wonderful Launcher v1.x
- ComfyUI: Portable / Managed install
- Python: 3.11+
- CUDA / Torch: CUDA 12.x / Torch 2.x
- Last tested: 2026-05-19
If you searched where to put safetensors in comfyui, where do safetensors go in comfyui, or comfyui safetensors folder, the answer is: put the file under ComfyUI/models/, but choose the subfolder by model type, not by the .safetensors extension.
Use the folder map first. Checkpoints go in models/checkpoints/, LoRA files go in models/loras/, VAE files go in models/vae/, Flux diffusion models usually go in models/diffusion_models/, and text encoders usually go in models/text_encoders/ or models/clip/.
Most .safetensors files go somewhere under:
ComfyUI/models/If the file is in the wrong folder, ComfyUI may start normally but the model dropdown will be empty or the workflow will fail validation.
Quick Folder Map
| Model Type | Put It Here | Common Loader |
|---|---|---|
| Checkpoint | ComfyUI/models/checkpoints/ | Load Checkpoint |
| LoRA | ComfyUI/models/loras/ | Load LoRA |
| VAE | ComfyUI/models/vae/ | Load VAE |
| ControlNet | ComfyUI/models/controlnet/ | Load ControlNet Model |
| Embedding/Textual Inversion | ComfyUI/models/embeddings/ | Used in prompt text |
| CLIP/Text encoder | ComfyUI/models/text_encoders/ or models/clip/ | CLIP loader nodes |
| Flux diffusion model / UNet | ComfyUI/models/diffusion_models/ or models/unet/ | UNet/Diffusion model loader |
| Upscaler | ComfyUI/models/upscale_models/ | Upscale Model Loader |
After moving files, restart ComfyUI or refresh the model list from the relevant dropdown.
If you only remember one rule, remember this one: the .safetensors extension does not tell you the right folder by itself. checkpoint, LoRA, VAE, ControlNet, Flux diffusion model, and text encoder can all use .safetensors while belonging in different places.
Common File Examples
| File or Name Pattern | Usually Put It Here | Why |
|---|---|---|
sd_xl_base_1.0.safetensors, realvisxl.safetensors | models/checkpoints/ | Full checkpoint loaded by Load Checkpoint |
ae.safetensors for Flux workflows | models/vae/ | Flux autoencoder / VAE component |
umt5_xxl_fp8_e4m3fn_scaled.safetensors | models/text_encoders/ | Text encoder used by newer Flux-style workflows |
qwen_*.safetensors text encoder files | models/text_encoders/ | Text encoder, not a checkpoint |
4x_*.safetensors or ESRGAN-style upscalers | models/upscale_models/ | Used by Upscale Model Loader |
| Flux GGUF or standalone diffusion model files | models/diffusion_models/ | Loaded through a diffusion, UNet, or GGUF loader |
When a filename includes a model family such as Flux or Qwen, do not assume it belongs in models/checkpoints/. Check the loader node in the workflow first.
Fast Diagnosis
| What You See | What It Usually Means | First Thing to Check |
|---|---|---|
| The model file never appears in the dropdown | Wrong folder | Match the file type to the table above |
| Flux or UNet file is missing from Load Checkpoint | Wrong loader and wrong folder | Move it to diffusion_models and use a diffusion model loader |
| LoRA file is in the checkpoint list nowhere | It is not a checkpoint | Put it in models/loras and use Load LoRA |
| Text encoder file is not detected | Wrong CLIP/text encoder location | Check models/text_encoders or models/clip |
| File is in the right folder but still missing | Wrong ComfyUI install path or stale file list | Restart ComfyUI and confirm you are editing the same install that is running |
The Extension Is Not the Model Type
This is the root of most where to put safetensors confusion.
These files can all end with .safetensors, but they do not belong in the same folder:
- a full checkpoint
- a LoRA
- a VAE
- a ControlNet model
- a Flux diffusion model
- a text encoder
If you only sort by file extension, you will eventually put the right file in the wrong folder.
Checkpoints
Put standard Stable Diffusion or SDXL checkpoint files here:
ComfyUI/models/checkpoints/Examples:
realvisxl.safetensors
dreamshaper.safetensors
sd_xl_base_1.0.safetensorsUse the Load Checkpoint node. If ckpt_name says the model is not in the list, confirm the file is in models/checkpoints/ and not in models/loras/ or models/diffusion_models/.
LoRA Files
Put LoRA files here:
ComfyUI/models/loras/Use a Load LoRA node. A LoRA is not a full checkpoint. It needs a compatible base model.
If the LoRA was trained for SD 1.5, use it with an SD 1.5 checkpoint. If it was trained for SDXL, use it with an SDXL checkpoint.
VAE Files
Put VAE files here:
ComfyUI/models/vae/Use Load VAE or the VAE slot in a workflow. A wrong VAE can cause color, decode, or model mismatch problems.
ControlNet Files
Put ControlNet files here:
ComfyUI/models/controlnet/ControlNet files must match the model family. SD 1.5 ControlNet models are not interchangeable with SDXL ControlNet models.
Flux, UNet, and Diffusion Models
Many newer workflows, including Flux-style workflows, do not use a single checkpoint file in the old SD 1.5 pattern.
For standalone diffusion model files, use:
ComfyUI/models/diffusion_models/Some older workflows or custom nodes may still refer to:
ComfyUI/models/unet/ComfyUI maps legacy unet paths to diffusion_models, so both patterns can appear in guides and custom nodes.
If your file is a Flux-style model and you keep checking Load Checkpoint, it can look like the model disappeared. In reality, you are usually checking the wrong loader for that file type.
GGUF Models
GGUF files are usually quantized model files for lower VRAM workflows. With ComfyUI-GGUF, Flux GGUF diffusion models commonly go in:
ComfyUI/models/diffusion_models/Many ComfyUI-GGUF examples also use the legacy-compatible folder:
ComfyUI/models/unet/Then use Unet Loader (GGUF) or the GGUF loader required by the workflow. A GGUF diffusion model is not a normal checkpoint.
If you searched for comfyui unet loader gguf or comfyui gguf loader, check both the model folder and whether the ComfyUI-GGUF custom node is installed. See ComfyUI GGUF for the full loader setup.
Text Encoders and CLIP
Put text encoder files in:
ComfyUI/models/text_encoders/Some workflows and older guides use:
ComfyUI/models/clip/ComfyUI supports both locations for text encoders in current folder path configuration.
Why the Dropdown Is Still Empty
If the file is in the right folder but does not appear:
- Check the extension.
.safetensorsis supported, but a partially downloaded file may not be. - Refresh the dropdown or restart ComfyUI.
- Make sure the file is not inside a zip or browser download temp file.
- Confirm you are editing the same ComfyUI folder you are running.
- Check
extra_model_paths.yamlif models live outside the ComfyUI folder.
What Not to Do
Avoid these common mistakes:
- putting every
.safetensorsfile inmodels/checkpoints - using Load Checkpoint for a Flux diffusion model
- putting LoRA files next to full checkpoints
- editing one ComfyUI install while launching a different one
- assuming a browser refresh is enough after moving many models
Portable Package Example
For a portable install:
ComfyUI_windows_portable/
ComfyUI/
models/
checkpoints/
loras/
vae/
controlnet/
diffusion_models/If your launch .bat starts a different ComfyUI folder, the model will not show up. Always verify the folder shown in the running terminal or launcher settings.
How Wonderful Launcher Helps
Wonderful Launcher is useful when model placement is mixed with environment problems:
- preserve model folders before rebuilding ComfyUI
- keep stable and experimental model libraries separate
- avoid losing workflows while fixing missing custom nodes
- collect logs when model dropdowns or workflow validation fail
If the model file is correct but the workflow uses missing custom nodes, start with Install Custom Nodes.
If the workflow still refuses to open after the model is in the right place, save the workflow JSON before changing more files. See How to Save a Workflow in ComfyUI.
FAQ
Where do .safetensors files go in ComfyUI?
They usually go somewhere under ComfyUI/models/, but the exact folder depends on whether the file is a checkpoint, LoRA, VAE, ControlNet model, Flux diffusion model, or text encoder.
Can I put all .safetensors files in models/checkpoints?
No. That is the most common mistake. Many .safetensors files are not checkpoints at all.
Where do Flux .safetensors models go in ComfyUI?
Usually in ComfyUI/models/diffusion_models/ or the legacy models/unet/ path, then loaded with the appropriate diffusion model or UNet loader.
Why is my .safetensors file not showing up?
Usually because it is in the wrong folder, you are looking at the wrong loader, or you edited a different ComfyUI install than the one that is currently running.
Related Guides
- Download Models
- ComfyUI GGUF
- ComfyUI GGUF and model loading basics
- Install Custom Nodes
- ComfyUI Common Issues
Source References
Start with Wonderful Launcher if this issue touches your real ComfyUI environment. Use the docs to understand the fix, and use the app to inspect the machine you already have.
Download Wonderful LauncherStart with Wonderful Launcher if this issue touches your real ComfyUI environment. Use the docs to understand the fix, and use the app to inspect the machine you already have.
Download Wonderful LauncherDid this fix your issue?
Your answer helps prioritize verified ComfyUI repairs.
Manager Install Failed
How to fix deployment failures at the ComfyUI-Manager installation step, especially on Windows networks where GitHub clone, proxy, SSL, or path issues block first-run setup.
Workflow Missing Nodes
Fix ComfyUI workflows that fail to load because of missing custom nodes, with steps to identify, install, and recover missing node types.