Stability AI
Updated May 2026
Stable Diffusion
The original open-source image model
Stability AI released Stable Diffusion in 2022 and single-handedly created the open-source image generation ecosystem. The company went through near-bankruptcy and multiple leadership changes, but the models keep coming. SD 3.5 in late 2024 returned to frontier quality. Today SD is the tinkerer's choice: runs on a gaming GPU, thousands of community fine-tunes and LoRAs, and powers Automatic1111 and ComfyUI.
Variants
4
Best at
truly open weights with permissive community license
Max context
—
Pricing
Self-host
$0
per community license
Stability AI API
~$0.04
per image (SD 3.5 Large)
Variants
Sort the table by context window or cost to find the right variant. Click any version below for a battle card with ranks, pricing notes, and official links.
| Modalities | ||||
|---|---|---|---|---|
Stable Diffusion 3.5 Large sd-3-5-large | — | varies / varies | 2024 | image |
Stable Diffusion 3.5 Medium sd-3-5-medium | — | varies / varies | 2024 | image |
Stable Audio 2 stable-audio-2 | — | varies / varies | 2024 | audio |
Stable Video Diffusion stable-video | — | varies / varies | 2024 | video |
Battle card
Context rank
N/A
within Stable Diffusion
Capability rank
#1
modalities + reasoning
Weights
Open
self-hostable if licensed
Best fights to pick
- open-source art generation
- heavy fine-tuning
- offline image workflows
Rankings are Tendril directory ranks, computed from the model data shown here. Public benchmark leaderboards change often, so official docs and current benchmark pages should be checked before buying or deploying.
Learn
Lessons about this model
Structured lessons that cover Stable Diffusion directly or put it in context alongside its rivals.
Check yourself
Quizzes
Short, mixed-difficulty quiz sets on Stable Diffusion and its model family.
Open-Weight Families: Llama, Mistral, Qwen, DeepSeek, Gemma
7 questions
The open ecosystem that shook the industry.
Start quiz →Hands-on
Try these prompts
Ready-made prompts that show Stable Diffusion at its best. Use them in your own AI workspace, then compare the output with what you learned in Tendril.
SDXL Turbo local prompt
CreatorsSDXL Turbo generates in under a second on a decent GPU — use ComfyUI or Automatic1111 locally.
cinematic still, rainy tokyo alleyway at night, neon reflections on wet pavement, single figure with an umbrella, 35mm, shallow depth of field
LoRA fine-tune on your style
CreatorsStable Diffusion's superpower is you can train a LoRA on 15-20 of your images.
<lora:my-illustration-style:0.85> a friendly cartoon fox reading a book in a library, my illustration style, warm palette
ControlNet for pose control
CreatorsConstrain the pose with a skeleton image — SD's killer-feature combo.
[Pose control image: running figure] photorealistic portrait of a marathon runner crossing the finish line, overcast day, city background
Print & keep
Printable reference
One-page summaries and flowcharts — great for desks, classrooms, or study sessions.
Go deeper
Official resources
Straight from the lab — docs, API references, and the chat surfaces you can try today.
Strengths
- truly open weights with permissive community license
- enormous fine-tune ecosystem
- runs on consumer hardware
Limits
- Stability AI's finances have been rocky
- quality trails FLUX and Midjourney
- license gets stricter above $1M revenue