# Stable Diffusion XL base 1.0 — card for AI agents

Stability AI's open text-to-image latent diffusion model; works alone or with the SDXL refiner for final detail.

Category: AI image generators · maker Stability AI · checked 2026-10-25

## Price
- Price: Free
- Free tier: Open weights, free to download.

## License
CreativeML Open RAIL++-M — https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/LICENSE.md

- commercial: **conditions** (Commercial use with conditions)
  > Therefore You cannot use the Model and the Derivatives of the Model for the specified restricted uses.
  (https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/LICENSE.md)
- attribution: **no** (No credit needed)
  > You must give any Third Party recipients of the Model or Derivatives of the Model a copy of this License
  (https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/LICENSE.md)
- in app: **conditions** (Ship inside your app with conditions)
  > You may host for Third Party remote access purposes (e.g. software-as-a-service), reproduce and distribute copies of the Model or Derivatives of the Model
  (https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/LICENSE.md)
- youtube: **conditions** (Monetized YouTube with conditions)
  > Licensor claims no rights in the Output You generate using the Model.
  (https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/LICENSE.md)

## Rules to follow
- Follow the use-based restrictions in Attachment A of the license.
- Pass those restrictions on to your users if you host or redistribute the model.
- Ship a copy of the license with any redistributed weights or derivatives.

## How to connect
- Download only
- Repository: https://github.com/Stability-AI/generative-models
- Website: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0

## Notes
Hugging Face API, Oct 25, 2026: 4,204,941 downloads, 8,392 likes; GitHub Stability-AI/generative-models: 27,312 stars.

## Numbers (AgentGid data)
- Gid Score: 44
- GitHub stars: 27,345

## Lines to paste into your prompt or AGENTS.md

```
- Stable Diffusion XL base 1.0 (AI image generators). Stability AI's open text-to-image latent diffusion model; works alone or with the SDXL refiner for final detail. Price: free; commercial use with conditions; no credit needed; ship inside your app with conditions.
  Connect: download from https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0 (no API).
  Rule: Follow the use-based restrictions in Attachment A of the license.
  Rule: Pass those restrictions on to your users if you host or redistribute the model.
  Rule: Ship a copy of the license with any redistributed weights or derivatives.
  License: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/LICENSE.md (checked 2026-10-25; details: https://agentgid.com/assets/sdxl/)
```

## Links
- AgentGid page: https://agentgid.com/assets/sdxl/
- Is it free for commercial use? https://agentgid.com/assets/sdxl/commercial-use/
- Source: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0
- Source: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/LICENSE.md
- Source: https://huggingface.co/api/models/stabilityai/stable-diffusion-xl-base-1.0
- Source: https://github.com/Stability-AI/generative-models


## SDXL base 1.0: open model that draws pictures from text

Stable Diffusion XL is a free, open image model from Stability AI that turns a written description into a picture. You download it and run it on a computer with a strong graphics card, or inside an app that runs it for you. Its license allows commercial use, but forbids a list of harmful uses that you must pass on to your own users.

### Where to use it
- Artwork for design work: Make illustrations, mood boards or background art from a short description while you design a page, a poster or a game.
- A picture feature in a learning app: Let students or users of a creative app turn their own words into a picture, with the model running on your server.
- Restyle a sketch or photo: Start from an existing picture and a description, and get a new version that keeps the layout, for example a rough sketch turned into a finished scene.
- Replace one part of an image: Mark an area with a mask and describe what should be there, for example swapping the person in a stock-style scene for a diver.
- Extra detail with the refiner: Run a second, companion model after the base one to add fine detail to the final picture, at the cost of more time and memory.
- Placeholder art for a prototype: Fill a game or app prototype with quick concept pictures so testers see a finished look before an artist draws the real ones. (idea, not from the official pages)

### How to start
- Install Diffusers 0.19 or newer with Transformers, Accelerate, safetensors and the invisible watermark package.

```
pip install invisible_watermark transformers accelerate safetensors
```

- Load the base model in half precision and generate an image from a prompt on an Nvidia card.

```
pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, use_safetensors=True, variant="fp16")
```

- With little graphics memory, move parts to the main memory instead of putting the whole model on the card.

```
pipe.enable_model_cpu_offload()
```


### License
You may use it, host it for others and share it, also commercially, under CreativeML Open RAIL++-M. A list of harmful uses is forbidden, and you must pass that list on to your users.

### Good to know
- The model page itself says it is meant for research, while the license allows wider use with limits. Read both before a commercial launch.
- It cannot write readable words inside pictures, and faces or people may come out wrong.
- Forbidden uses include harming minors, harassment, false information meant to harm, medical advice and automated decisions that hurt people's rights.
- Pictures come out best at 1024 by 1024; sizes below 512 by 512 are unlikely to work.
- The download is large; the single-file version of the base model alone is about 6.9 GB.

Checked 2026-11-10 against: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0, https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/raw/main/README.md, https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/LICENSE.md, https://huggingface.co/api/models/stabilityai/stable-diffusion-xl-base-1.0/tree/main, https://huggingface.co/docs/diffusers/main/en/using-diffusers/sdxl, https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/stable_diffusion_xl, https://github.com/Stability-AI/generative-models, https://pypi.org/pypi/diffusers/json

### Mistakes to avoid
- Don't: Generating at 512x512 or smaller to save time → Do: Keep the default 1024x1024, or 768x768 at the least (The model was built for 1024 pixels and small sizes break down.)
- Don't: pipe.to("cuda") on a card with too little memory → Do: pipe.enable_model_cpu_offload() instead of .to("cuda") (Offloading moves idle parts to main memory so the model fits.)
- Don't: Prompting for a logo or a banner with exact words in it → Do: Generate the picture without text and add the words later in your design tool or page (The model cannot draw readable text.)
- Don't: Hosting SDXL for users with no mention of the license restrictions → Do: Put the Attachment A use restrictions into your terms and tell users the model is subject to them (The license requires the restrictions to be enforceable on everyone you serve.)

### Questions
- Can I sell images I make with SDXL? The license claims no rights in your output, but the use restrictions still apply to how you use it.
- Can I offer SDXL to my users as an online service? Yes, hosting for others is allowed if your terms include the use restrictions and you give users notice of them.
- Do I need the refiner model? No. The base model works alone; the refiner is an optional second step for fine detail.
- Are my images marked as AI-made? Diffusers adds an invisible watermark by default when the invisible-watermark package is installed.

---
AgentGid's catalog is curated: every entry was checked against the maker's official pages (registry, repository, license and pricing pages) on the date shown; a license value shows the quote it stands on wherever AgentGid has one. Usage numbers come from AgentGid's own scan of public repositories set up for AI coding agents. Claims by makers are reported, not tested by AgentGid.
