Stability AIBlack Forest Labs

Run FLUX and SDXL at home: which GPU do you need?

Published Oct 27, 2026 Updated Oct 27, 2026

We earn from qualifying purchases via affiliate links. Disclosure.

Verdict

Stability AI says SDXL 1.0 works on consumer GPUs with 8 GB of VRAM. FLUX.1 dev needs about 12.1 GB at FP8 or 7.3 GB as a GGUF Q4, per a third-party estimate (±20%). So a 12 GB card covers SDXL and FLUX at Q4, and a 16 GB card covers FLUX at FP8.

Prices: Newegg checked Oct 27, 2026, about 12:35 PM CT; Minisforum checked Oct 27, 2026, 12:40 PM CT. New condition only.

Model memory at a glance

Model	Memory needed	Format	Context	Smallest tier (our estimate)	Source
SDXL 1.0Stability AI · Image	8 GB VRAM	as published	—	8–12 GB GPU8–12 GB GPU	Stability AI: SDXL 1.0 announcement
FLUX.1 devBlack Forest Labs · Image	12.1 GB FP8; 7.3 GB GGUF Q4 (estimate)	FP8 / Q4	—	16 GB GPU16 GB GPU; 12 GB with Q4	vramcalculator.net: FLUX.1 dev VRAM estimate (third party)

Good, Better, Best for this model (8–12 GB tier)

Three new-retail picks, chosen from cited specs and prices. We do not test hardware ourselves. Prices show when we checked them.

Good

Cheapest 12 GB card we priced

ASRock Challenger Intel Arc B580 12GB

$329.99 Newegg · checked Oct 27, 2026, about 12:35 PM CT

Holds 8B–14B chat models at Q4 and SDXL (per the sources). Needs a desktop to go in; check your runtime supports Intel Arc.

Sold by Newegg, seen on the product page

Check price at Newegg

Better

Step up to 16 GB

ASUS Dual GeForce RTX 5060 Ti 16GB OC

Price shown at Amazon

4 GB more VRAM than the Good pick, which opens up gpt-oss 20B (14 GB download). Needs a desktop to go in.

Sold at Amazon.

Check price at Amazon

Best

ABS Cyclone series mesh mini-tower case with RGB fans — illustrated product card

Complete PC, nothing to build

ABS Cyclone Aqua gaming PC (Core i7-14700F, RTX 5060 Ti 16GB, 16GB DDR5, 1TB SSD)

$1,599.99 Newegg · checked Oct 27, 2026, about 12:35 PM CT

One box with the 16 GB card already inside. Ships with 16 GB system RAM; plan an upgrade to 32 GB.

Sold by Newegg, seen on the product page. Ships with 16 GB system RAM; plan an upgrade to 32 GB.

Check price at Newegg

The details

SDXL

Stability AI's SDXL 1.0 announcement says it "should work effectively on consumer GPUs with 8GB VRAM." That is the maker's figure, not ours.

FLUX.1 dev

We use vramcalculator.net's estimate rather than a maker figure: 12.1 GB at FP8 and 7.3 GB as a GGUF Q4, with that site's own ±20% caveat. Treat those as estimates.

Software support

Image tools vary in how well they support Intel and AMD cards. Check that your tool supports the card before buying; we have not tested any of these cards.

Install and run FLUX and SDXL

Steps and commands come from the official docs listed at the end of this section, checked Oct 27, 2026, 2:30 PM CT. We don't add flags the docs don't show. Commands can change, so check the linked page if one fails.

1. Before you start: GPU drivers

Install your GPU's current driver: NVIDIA (CUDA), AMD (ROCm on Linux, or a current Adrenalin driver on Windows 11), or Intel Arc. ComfyUI's manual install supports NVIDIA, AMD, Intel and Apple silicon.

2. Install ComfyUI

Easiest (Windows and macOS): the ComfyUI desktop app from comfy.org/download.

Windows portable builds: ComfyUI ships portable 7z builds for NVIDIA (20 series and newer) and Intel GPUs. Extract one and run it. Put checkpoints in ComfyUI\models\checkpoints.

Manual install (Windows, Linux): clone the repo, install PyTorch for your GPU, then install the requirements. Python 3.13 is the best-supported version.

git clone https://github.com/Comfy-Org/ComfyUI.git
cd ComfyUI

GPU	PyTorch command from the ComfyUI README
NVIDIA	`pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu130`
AMD on Linux (incl. Strix Halo)	`pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm7.2`
AMD on Windows 11 (incl. R9700 and Strix Halo)	Use AMD's multi-architecture PyTorch packages with a current AMD driver and 64-bit Python 3.13. The README lists the device extras, e.g. `device-gfx1201` for the AI PRO R9700 and `device-gfx1151` for Ryzen AI Max.
Intel Arc	`pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/xpu`
Apple silicon	Install PyTorch nightly following Apple's Accelerated PyTorch on Mac guide.

pip install -r requirements.txt
python main.py

3. Add SDXL or FLUX

Put checkpoints (the large .safetensors files) in models/checkpoints and VAE files in models/vae. Larger models like FLUX come as several files, so follow each model's own instructions for which subfolder each file goes in.

4. Check that it's running on the GPU

Watch GPU memory while an image renders:

nvidia-smi   # NVIDIA
rocm-smi     # AMD on Linux

5. Common fixes

"Torch not compiled with CUDA enabled": run pip uninstall torch, then reinstall with the NVIDIA command above.
Low VRAM: ComfyUI says it can run large models on as little as 4 GB of VRAM and 8 GB of RAM by streaming weights. It will be slower than fitting the model entirely in VRAM.
Strix Halo: on Windows, raise Variable Graphics Memory in AMD Software (Performance, then Tuning, then restart). On Linux, see AMD's Strix Halo memory notes (BIOS VRAM, or GTT size).