Stable Diffusion

weights

open(SDXL and SD-Turbo download without a gate

data

described(the SDXL and SD3 reports describe the training data

code

open(Stability-AI/generative-models publishes the training framework, with example training configs, under MIT)

license

Stability-AI-Community-License(SD3, SD3.5, SDXL-Turbo and SD-Turbo: free below USD 1M annual revenue, after which the license terminates and an enterprise license is required)+CreativeML-OpenRAIL++-M(SDXL 1.0 base and refiner: use-based restrictions only)

The Stable Diffusion 3.5 models and SDXL-Turbo are free to use, commercially included, for organizations under USD 1M in annual revenue; above that a paid license is required. SDXL itself carries only use-based restrictions, and the training framework is published while the training data is not. Stability-AI-Community-License allows commercial use only within a bound, so the license tier is use_bounded.

• https://cdn.jsdelivr.net/gh/Stability-AI/generative-models@main/LICENSE-CODE recorded 2026-09-26

  LICENSE-CODE body: "MIT License Copyright (c) 2023 Stability AI".

• https://cdn.jsdelivr.net/gh/Stability-AI/generative-models@main/README.md recorded 2026-09-26

  generative-models README: "For training, we use PyTorch Lightning", with example training configs.

• https://huggingface.co/api/models?author=stabilityai&sort=downloads&direction=-1&limit=100 recorded 2026-09-26

  Hugging Face listing of the stabilityai organization: the SDXL, SD-Turbo, SD3 and SD3.5 checkpoints are public repositories.

• https://huggingface.co/stabilityai/sd-turbo/raw/main/LICENSE.md recorded 2026-09-26

  SD-Turbo license body: "STABILITY AI COMMUNITY LICENSE AGREEMENT Last Updated: July 5, 2024".

• https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/raw/main/LICENSE.md recorded 2026-09-26

  SDXL license body: "CreativeML Open RAIL++-M License dated July 26, 2023", with use-based restrictions in Attachment A and no revenue threshold.

• https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/raw/main/README.md recorded 2026-09-26

  SDXL card: points to the SDXL report on arXiv and to generative-models, "(both training and inference)"; no training dataset is released.

• https://stability.ai/community-license-agreement recorded 2026-09-26

  Stability AI Community License: "this Agreement preserves free access to the Models for people or organizations generating annual revenue of less than US $1,000,000"; above it "any licenses granted to You under this Agreement shall terminate".

• https://stability.ai/license recorded 2026-09-26

  License page: "Community For researchers, developers, small businesses, and creators with less than $1M in annual revenue. License includes: Stable Audio 3.0 Stable Diffusion 3.5 Suite SDXL Turbo Stable Fast 3D".

Adoption is measured as Hugging Face downloads across the nine checkpoints on Stability AI's own account, most of them for SDXL base. The Stable Diffusion 1.x mirrors on other accounts are left out.

• https://huggingface.co/api/models?author=stabilityai&sort=downloads&direction=-1&limit=100 recorded 2026-09-26

  Organization listing sorted by downloads: stable-diffusion-xl-base-1.0 3598555, sdxl-turbo 990384, sd-turbo 388670, stable-diffusion-3.5-medium 115315 in the trailing 30 days.

Stable Diffusion 3.5 Large sits in the lower half of the text-to-image arena, well behind FLUX.2 [dev] and Qwen-Image among open models. It is placed and compared there, which newer entrants with no arena presence are not.

• https://artificialanalysis.ai/image/leaderboard/text-to-image recorded 2026-09-26

  Arena leaderboard entries "Stable Diffusion 3.5 Large" (rank 128, Elo 838.8), "Stable Diffusion XL 1.0" (rank 159) and "Stable Diffusion 1.5" (rank 163).