| File | flux2-vae.safetensors |
| What it is | FLUX.2 VAE (VAE) |
| Size | 0.3 GB |
| Folder | ComfyUI/models/vae/ |
| Loader node | Load VAE (VAELoader) |
| Source | Comfy-Org/flux2-dev on Hugging Face |
Download flux2-vae.safetensors (0.3 GB)
Where to put it
Put the file in ComfyUI/models/vae/ (in the portable version: ComfyUI_windows_portable/ComfyUI/models/vae/). Then press R in ComfyUI, or restart it, so the file shows up in Load VAE (VAELoader).
If a workflow says “Value not in list” for this node, the file name in the workflow is different from yours, or the file is in another folder. Click the node and pick the file from the list.
What it does
The VAE turns the finished latent into pixels (and pixels into latents, for image-to-image and editing). It is small, but each model family needs its own: a VAE from another family gives grey, noisy or coloured-noise images.
Models that use flux2-vae.safetensors
| Model | Type | Fits from | Runs well from |
|---|---|---|---|
| ERNIE-Image (Turbo) | image | 6 GB | 11 GB |
| FLUX.2 [dev] | image | 17 GB | 39 GB |
| FLUX.2 [klein] 4B | image | 4 GB | 6 GB |
| FLUX.2 [klein] 9B | image | 7 GB | 12 GB |
| Ideogram 4 | image | 15 GB | 21 GB |
VRAM of the graphics card, for the model with the file this site picks for that card. Open a model to see every file and every GPU.
Tested with this file
I use this exact file in 2 of my free workflows, measured on an RTX 5060 Ti 16 GB:
- FLUX.2 klein 4B — FP8 — 2.9 s per 1024×1024 image
- FLUX.2 klein 9B — GGUF Q8_0 — 8.5 s per 1024×1024 image
Not sure your card can run these models? Check your GPU. All shared files: ComfyUI model files.
Size and source read from Hugging Face (2026-10-01).