Stable Diffusion
Visit Stable Diffusion →About Stable Diffusion
Stable Diffusion is a family of open-weights image generation models from Stability AI. Unlike closed services, you can download the weights and run them on your own hardware, which is why it became the default base for a huge ecosystem of fine-tunes, LoRAs, ControlNets and local UIs.
The current release is Stable Diffusion 3.5, shipping as Large (8.1B parameters), Large Turbo for fast few-step generation, and Medium (2.5B) for lighter machines, all available as open weights. Stability's hosted API also lists a Stable Diffusion 3.5 Flash variant, though its open-weights status has not been confirmed.
Capabilities cover text-to-image generation, prompt adherence and typography that improved over SDXL, and a broad toolchain for control and editing thanks to the open-model community. Related open models from Stability include SDXL Turbo, Stable Audio 3.0 and Stable Fast 3D.
On pricing, self-hosting the open weights is free under the Stability AI Community License for individuals and organizations under $1M in annual revenue. Above that threshold an Enterprise License with custom pricing is required. Stability also offers a hosted route through the Developer Platform API (platform.stability.ai) with pay-per-credit billing at $1 per 100 credits (1 credit is about $0.01): Stable Image Ultra costs 8 credits (about $0.08 per image), SD3.5 Large 6.5 credits, Large Turbo 4 credits, Medium 3.5 credits, Flash 2.5 credits, and Core 3 credits (about $0.03 per image), with SDXL 1.0 starting from 0.9 credits. New API accounts receive some free starter credits.
The main limitations are practical. Running the larger models locally needs a capable GPU, results usually require prompt work and post-processing, and the revenue-gated license means larger companies cannot treat it as simply free.
Stable Diffusion is the right choice if you want local control, an open pipeline you can fine-tune, and no per-image cost at small scale, and you are comfortable prioritizing openness and flexibility over relying on a single hosted model.
- Open weights you can download and run locally with full control over the pipeline
- Free to self-host under the Community License for individuals and orgs under $1M revenue
- Three SD 3.5 variants (Large, Large Turbo, Medium) let you match the model to your hardware
- Enormous ecosystem of fine-tunes, LoRAs, ControlNets and local UIs built around it
- Optional hosted API with transparent per-image credit pricing if you don't want to run it yourself
- The larger models need a capable GPU to run smoothly at home
- Free self-hosting is revenue-gated; orgs over $1M need a paid Enterprise License
- Good results usually require prompt tuning and post-processing
- Hosted API costs add up at volume (about $0.08/image for Ultra, $0.03 for Core)
Choose Stable Diffusion if…
- You want to run an image model locally with full control
- You need an open base to fine-tune with LoRAs or ControlNets
- Your revenue is under $1M and you want zero per-image cost
- You prioritize openness and flexibility over relying on a single hosted model
- You want the option of a paid hosted API without vendor lock-in
Leave a Review
Reviews are published after moderation. We don't share your email.