| File | flux2_dev_fp8mixed.safetensors |
| Model | FLUX.2 [dev] (image model, 32B) |
| Precision | FP8 |
| Size | 35.5 GB |
| VRAM to fit | about 38.8 GB |
| Folder | ComfyUI/models/diffusion_models/ |
| Loader node | Load Diffusion Model (UNETLoader) |
| Source | Comfy-Org/flux2-dev on Hugging Face |
Download flux2_dev_fp8mixed.safetensors (35.5 GB)
Where to put it
Put the file in ComfyUI/models/diffusion_models/ (portable version: ComfyUI_windows_portable/ComfyUI/models/diffusion_models/). Older guides say models/unet: that folder still works. Press R in ComfyUI or restart it, then pick the file in Load Diffusion Model (UNETLoader). Not for Mac: FP8 and INT8 files do not load on Apple Silicon.
If a workflow says “Value not in list” for that node, the workflow expects a file with another name. Click the node and choose this one.
Which graphics cards it fits
| VRAM | This file | Best file for that card (RTX 40/50) | |
|---|---|---|---|
| 6 GB | Not practical | Short by 32.8 GB | flux2-dev-Q2_K.gguf (12.9 GB) |
| 8 GB | Not practical | Short by 30.8 GB | flux2-dev-Q2_K.gguf (12.9 GB) |
| 10 GB | Not practical | Short by 28.8 GB | flux2-dev-Q2_K.gguf (12.9 GB) |
| 12 GB | Not practical | Short by 26.8 GB | flux2-dev-Q4_K_M.gguf (20.1 GB) |
| 16 GB | Not practical | Short by 22.8 GB | flux2-dev-Q2_K.gguf (12.9 GB) |
| 20 GB | Offload only | Short by 18.8 GB: runs with part of it in system RAM, slower | flux2-dev-Q3_K_M.gguf (16.0 GB) |
| 24 GB | Offload only | Short by 14.8 GB: runs with part of it in system RAM, slower | flux2-dev-Q4_K_M.gguf (20.1 GB) |
| 32 GB | Offload only | Short by 6.8 GB: runs with part of it in system RAM, slower | flux2-dev-Q6_K.gguf (27.4 GB) |
| 48 GB | Runs well | Fits in VRAM, 9.2 GB to spare | this file |
Need = file × 1 + 2.5 GB for the picture and 0.8 GB for Windows and the browser — the same rule as the rest of the site. FP8 files do not load on a Mac (Apple Silicon); use a 16-bit or GGUF file there.
What else you need
This file alone does not make a picture. With it you load:
| Role | File | Size | Folder |
|---|---|---|---|
| Model | flux2_dev_fp8mixed.safetensors | 35.5 GB | models/diffusion_models |
| Text encoder | mistral_3_small_flux2_fp8.safetensors | 18.0 GB | models/text_encoders |
| VAE | flux2-vae.safetensors | 0.3 GB | models/vae |
53.8 GB in total. Smaller text-encoder versions and other options are on the FLUX.2 dev page.
Other files of FLUX.2 dev
| File | Precision | Size | VRAM to fit |
|---|---|---|---|
| flux2-dev.safetensors | 16-bit | 64.4 GB | 67.7 GB |
| flux2_dev_fp8mixed.safetensors | FP8 | 35.5 GB | 38.8 GB |
| flux2-dev-Q8_0.gguf | Q8_0 | 35.0 GB | 38.3 GB |
| flux2-dev-Q6_K.gguf | Q6_K | 27.4 GB | 30.7 GB |
| flux2-dev-Q5_K_M.gguf | Q5_K_M | 24.1 GB | 27.4 GB |
| flux2-dev-Q4_K_M.gguf | Q4_K_M | 20.1 GB | 23.4 GB |
| flux2-dev-Q3_K_M.gguf | Q3_K_M | 16.0 GB | 19.3 GB |
| flux2-dev-Q2_K.gguf | Q2_K | 12.9 GB | 16.2 GB |
Same model, different compression. 8-bit files (FP8, INT8, Q8_0) look almost the same as 16-bit; below Q5 the loss starts to show.
Want the answer for your exact card? Check your GPU. Every file of every model: ComfyUI model files.
Size and source read from Hugging Face (2026-10-01).