⬡ FLUX Loader¶
Load a FLUX architecture model with dual text encoders (CLIP-L + T5-XXL) and VAE.
Inputs¶
| Name | Type | Required | Description |
|---|---|---|---|
diff_model |
COMBO |
✅ | FLUX diffusion model (e.g. flux1-dev-fp8.safetensors or GGUF variant) |
clip_1 |
COMBO |
✅ | First text encoder — typically CLIP-L (clip_l.safetensors) |
clip_2 |
COMBO |
✅ | Second text encoder — typically T5-XXL (t5xxl_fp16.safetensors or GGUF) |
vae |
COMBO |
✅ | VAE model (e.g. ae.safetensors) |
Outputs¶
| Name | Type | Description |
|---|---|---|
model |
MODEL |
FLUX diffusion model |
clip |
CLIP |
Dual text encoder (CLIP-L + T5-XXL) |
vae |
VAE |
Autoencoder |
model_name |
STRING |
Diffusion model filename |
GGUF Support¶
The FLUX Loader automatically detects .gguf files and routes them through the built-in GGUF loader. You can mix formats:
- Model:
.safetensorsor.gguf(Q4_K_M, Q8_0, etc.) - CLIP:
.safetensorsor.gguf(individually — you can use GGUF T5 with safetensors CLIP-L)
VRAM Recommendations
| Model Format | VRAM Required |
|---|---|
| FLUX Dev FP16 | ~24 GB |
| FLUX Dev FP8 (e4m3fn) | ~12 GB |
| FLUX Dev GGUF Q4_K_M | ~6 GB |