Stable Diffusion
Stability AI
Stable Diffusion is an image generation model line from Stability AI. The latest release is Stable Diffusion 3.5 Large (29 Oct 2024). This page tracks all 9 Stable Diffusion releases since 22 Aug 2022, with specs, licensing, and how to access each one. Official site →
Stable Diffusion 3.5 Largereleased
Improved prompt adherence and image quality vs SD 3 Medium; also released Turbo variant
- License
- Stability AI Community License
- Availability
- Self-host, Community API
- Parameters
- 8.1B
Same commercial use restrictions as SD 3 Medium
Stable Diffusion 3.5 Mediumreleased
Runs on consumer hardware; same family as SD 3.5 Large
- License
- Stability AI Community License
- Availability
- Self-host, Community API
- Parameters
- 2.5B
- Architecture
- MMDiT-X
Stable Diffusion 3.5 Large Turboreleased
Distilled few-step variant of SD 3.5 Large
- License
- Stability AI Community License
- Availability
- Self-host, Community API
Stable Diffusion 3 Mediumreleased
First SD3 open-weight release; commercial use restricted for revenue >$1M
- License
- Stability AI Community License
- Availability
- Self-host, Community API
- Parameters
- 2B
- Architecture
- Multimodal Diffusion Transformer (MMDiT)
License allows commercial use below $1M annual revenue; larger businesses require enterprise license
Stable Diffusion XL 1.0released
Base + refiner two-stage pipeline; major quality jump over 2.1
- License
- OpenRAIL++-M (SDXL)
- Commercial use
- Allowed
- Availability
- Self-host, Community API
- Parameters
- 6.6B
- Resolution
- 1024x1024
Stable Diffusion 2.1released
Reduced NSFW filter strictness vs 2.0; improved photorealism
- License
- CreativeML OpenRAIL++-M
- Commercial use
- Allowed
- Availability
- Self-host, Community API
Stable Diffusion 2.0released
New text encoder; improved quality but broke many v1.x fine-tunes
- License
- CreativeML OpenRAIL++-M
- Commercial use
- Allowed
- Availability
- Self-host, Community API
- Resolution
- 768x768
- Text Encoder
- OpenCLIP ViT-H/14
Stable Diffusion 1.5released
Fine-tuned from 1.0 with additional steps; community workhorse for several years
- License
- CreativeML Open RAIL-M
- Commercial use
- Allowed
- Availability
- Self-host, Community API
Stable Diffusion 1.0released
- License
- CreativeML Open RAIL-M
- Commercial use
- Allowed
- Availability
- Self-host, Community API
- Resolution
- 512x512
- Architecture
- Latent Diffusion Model
FAQ
- When was Stable Diffusion 3.5 Large released?
- Stable Diffusion 3.5 Large was released on 29 Oct 2024.
- Is Stable Diffusion open source?
- Yes. Stable Diffusion 3.5 Large ships open weights under the Stability AI Community License licence.
- Where can I use Stable Diffusion?
- Stable Diffusion 3.5 Large is available via Self-host and Community API.
See Stable Diffusion in context on the image model timeline, or browse all model families.