What hardware do I need to run Stable Diffusion XL?

At its most compressed (FP16) it needs roughly 7GB of VRAM, and about 12GB for good quality. VRAM figures are approximate and depend on context length and settings.

Is Stable Diffusion XL free for commercial use?

Commercial use is permitted, but with conditions. SDXL is released under a CreativeML Open RAIL++-M licence, which permits commercial use but attaches use-based restrictions: you may not use the model for a list of prohibited purposes, and you must pass those restrictions on to anyone you share it with. Read the licence before relying on it at scale.

Can I run Stable Diffusion XL on Apple Silicon?

Yes. Stable Diffusion XL has builds optimised for Apple Silicon, through MLX or GGUF on a Mac.

Stable Diffusion XL

Name: Stable Diffusion XL
Author: Stability AI

Stability AI · Image generation · 3.5B · Released 26 July 2023

Permitted with conditions TextImage Open weights Apple Silicon

SDXL is the workhorse of local image generation, and its staying power comes less from the base model than from what surrounds it. Years of community fine-tunes, LoRAs, ControlNets, and tooling mean there is an SDXL checkpoint for almost any style, and it runs on hardware from an 8GB card upwards. It generates native 1024px images and is well supported across ComfyUI, InvokeAI, and Fooocus. Newer models like FLUX produce more photorealistic results and far better text, but nothing matches SDXL's depth of ecosystem and customisability, which keeps it the default for people who want control.

Strengths

Enormous ecosystem of fine-tunes, LoRAs, and tooling
Native 1024x1024 output, and light enough for an 8GB card
Mature and extremely well documented

Weaknesses

Weaker at rendering readable text than FLUX or Qwen-Image
The base model rewards careful prompting and often a refiner step
The RAIL licence carries use-based restrictions to be aware of

Hardware requirements

Quantisation	Approx. VRAM	Notes
FP16	~7GB	The usual format, fits an 8GB card, more headroom helps

Optimised builds available for Apple Silicon.

What you'd need to run this

Roughly what a machine to run this would need, at up to three levels of quality. Memory is the deciding factor.

Minimum to run it

FP16 · ~7GB needed

One 12GB GPU

NVIDIA GeForce RTX 3060 12GB

or a Mac or mini-PC with unified memory, if you prefer no discrete GPU, Mac mini M5 Pro .

At least 32GB of system RAM alongside the card.

around £700–£1,100

What else 12GB runs →

Licence

CreativeML Open RAIL++-M — read the licence

Stable Diffusion XL: common questions

What hardware do I need to run Stable Diffusion XL?: At its most compressed (FP16) it needs roughly 7GB of VRAM, and about 12GB for good quality. VRAM figures are approximate and depend on context length and settings.
Is Stable Diffusion XL free for commercial use?: Commercial use is permitted, but with conditions. SDXL is released under a CreativeML Open RAIL++-M licence, which permits commercial use but attaches use-based restrictions: you may not use the model for a list of prohibited purposes, and you must pass those restrictions on to anyone you share it with. Read the licence before relying on it at scale.
Can I run Stable Diffusion XL on Apple Silicon?: Yes. Stable Diffusion XL has builds optimised for Apple Silicon, through MLX or GGUF on a Mac.

Availability

Recommended for

Customisable image generation with community fine-tunes and LoRAs
Running on modest hardware, from 8GB upward
A well-trodden starting point for local image generation

Related models

Run it with

Related guides

Local image generation

Glossary

Catalogue entry last verified 30 July 2026. Specifications change; verify anything you are about to spend money on.