image model · flux · Windows
Can I run FLUX.1 dev on AMD Ryzen AI Halo (128GB)?
Yes. FLUX.1 dev runs on AMD Ryzen AI Halo (128GB) at Q4 GGUF (~6.5 GB of ~96 GB usable).
Runs at Q4 GGUF using ~6.5 GB of ~96 GB usable.
- Peak VRAM
- ~6.5 GB
- Usable on device
- ~96 GB
- Device memory
- 128 GB
- Quant
- Q4 GGUF
How to run it
Use ComfyUI or Draw Things at Q4 GGUF. The big text encoder is loaded to encode your prompt, then offloaded before generation, which is why peak VRAM stays near the backbone size rather than the sum of every file.
- Type
- image (DIT)
- Parameters
- 12B
- Peak VRAM
- ~6.5 GB at Q4 GGUF
- Resolution
- 1024×1024
- License
- FLUX.1-dev Non-Commercial License
- Memory
- 128 GB unified
- Usable for weights
- ~96 GB
- Power draw
- ~120 W
- Best runtime
- llama.cpp (Vulkan/ROCm) / LM Studio
You could also run
Run FLUX.1 dev on other hardware
FAQ
Can AMD Ryzen AI Halo (128GB) run FLUX.1 dev?
Yes. FLUX.1 dev runs on AMD Ryzen AI Halo (128GB) at Q4 GGUF (~6.5 GB of ~96 GB usable).
How much VRAM does FLUX.1 dev need?
AMD Ryzen AI Halo (128GB) has room to spare. At Q4 GGUF the realistic peak is ~6.5 GB of VRAM, versus ~33 GB with every component kept resident (no offload). With aggressive CPU offload it drops to ~3 GB, much slower.
What do I use to run FLUX.1 dev locally?
FLUX.1 dev runs in ComfyUI or Draw Things (among others). It loads as a diffusion checkpoint plus its text encoder and VAE, not a single chat command.
Sources
- amd.com/en/blogs/2025
- amd.com/en/blogs/2026
- cnx-software.com
- frame.work
- huggingface.co/black-forest-labs/FLUX.1-dev
- huggingface.co/black-forest-labs/FLUX.1-dev/blob
- huggingface.co/city96/FLUX.1-dev-gguf
- huggingface.co/city96/FLUX.1-dev-gguf/discussions
- huggingface.co/docs
- lmstudio.ai
- microcenter.com
- notebookcheck.net
- ollama.com
VRAM figures are sourced peak-usage anchors at the noted quant, validated 2026-08-03. See methodology.