All vikis
007 First Light
Aion 2
Ananta
ARC Raiders
ArcheAge Chronicles
Call of Duty: Modern Warfare 4
ChatGPT
Chrono Odyssey
Civilization 3
Civilization 4
Claude
Crimson Desert
Deadlock
Diablo 4
Diablo 5
Enshrouded
Fable
Final Fantasy Resonance
Football Manager 27
Forza Horizon 6
God of War
Grand Theft Auto 6
Heroes of Might and Magic 3Stable Diffusion
Stable Diffusion is Stability AI's family of open image models (SD 1.5, SDXL, SD 3.5) that anyone can download and run on their own computer. SD 3.5 is free to use commercially for those under $1 million in annual revenue, and Stability's API charges from $0.025 an image.
Stable Diffusion is the family of open-weight image models from Stability AI that made it possible to generate AI images on your own computer: the main versions are SD 1.5 (2022), SDXL (Jul 26, 2023) and Stable Diffusion 3.5 (Oct 22, 2024), still the newest Stable Diffusion image model as of Nov 6, 2026. The weights are free to download, SD 3.5 is free for commercial use under $1 million a year in revenue, and Stability's own API charges $0.025 to $0.08 an image.
Use Stable Diffusion when you want to run image generation offline and privately, fine-tune your own style or character (LoRAs), and use the huge library of community add-ons (ControlNet, custom checkpoints) in tools like ComfyUI. It now trails newer models on quality: SD 3.5 Large ranked 83rd of 84 on the Arena text-to-image board (updated Nov 2, 2026). For a newer open model try Flux (Klein) or Qwen's image models; for the best results without setup, GPT Image, Nano Banana or Midjourney. See Run AI Locally and Best AI Image Generator.
Quick facts
| Maker | Stability AI |
| First public release | Aug 22, 2022 |
| Latest image model | Stable Diffusion 3.5 (Large, Large Turbo, Medium), Oct 2024 |
| Get it | Weights on Hugging Face; Stability API; apps such as ComfyUI |
| Price | Free to download; API $0.025 to $0.08 an image (as of Nov 6, 2026) |
| License | Stability AI Community License for SD 3.5 (free under $1M revenue); OpenRAIL for SD 1.x and SDXL |
| Hardware | SD 3.5 Medium needs about 9.9 GB of VRAM |
| Open weights | Yes |
Versions
| Model | Released | Size | License |
|---|---|---|---|
| Stable Diffusion 1.x | Aug 22, 2022 | Ran in about 6.9 GB of VRAM | CreativeML OpenRAIL-M, commercial use allowed |
| SDXL 1.0 | Jul 26, 2023 | 3.5B base, 6.6B with refiner | CreativeML OpenRAIL++-M |
| SD 3.5 Large / Large Turbo | Oct 22, 2024 | 8.1B | Stability AI Community License |
| SD 3.5 Medium | Oct 29, 2024 | 2.5B | Stability AI Community License |
Stability AI's news page, read Nov 6, 2026, shows no new Stable Diffusion image model since 3.5; its 2025 and 2026 releases were audio models (Stable Audio 3.0, May 2026), a brand production platform, Brand Studio (Apr 2026), and enterprise deals.
License and prices
The Stability AI Community License lets researchers, developers, small businesses and creators with under $1 million in annual revenue use SD 3.5 free, commercially included, and you own your outputs. Above $1 million, a paid Enterprise License is needed. Fine-tunes and LoRAs are allowed. SD 1.5 and SDXL use the older OpenRAIL licenses, which allow commercial use with use restrictions.
The paid API, billed at 1 cent a credit, costs per image as of Nov 6, 2026 (Stability pricing):
| Model | Price an image |
|---|---|
| Stable Image Ultra | $0.08 |
| SD 3.5 Large | $0.065 |
| SD 3.5 Large Turbo | $0.04 |
| SD 3.5 Medium | $0.035 |
| Stable Image Core | $0.03 |
| SD 3.5 Flash | $0.025 |
How to run Stable Diffusion locally
You need the model weights (from Hugging Face, after accepting the license) and an app to run them. The two most used are ComfyUI, a node-based editor where you wire steps into a workflow, and AUTOMATIC1111's Stable Diffusion web UI, a simpler form-style interface; both are free and open source. An NVIDIA graphics card is the easy path; Stability has also published versions optimized for AMD Radeon GPUs and NVIDIA TensorRT. SD 1.5 and SDXL run on modest cards, while SD 3.5 Medium wants about 10 GB of VRAM. The same apps also run Flux and other open models, which many users now load instead. See Run AI Locally and, for local text models, Ollama and LM Studio.
Getty Images lawsuit
Getty Images sued Stability AI in England over the use of its photos to train Stable Diffusion. In the UK, the High Court ruled on Nov 4, 2025 that Getty's secondary copyright claim failed: in the judge's words, a model that does not store or reproduce the works is not an "infringing copy". Getty won in part on trademarks, because early versions (v1.x and v2.x) could generate Getty and iStock watermarks, but the judge called those findings historic and "extremely limited in scope", and found no such evidence for SDXL. The judge made no finding on how many Getty works were used in training, and said Getty might bring such a case where the model was actually trained.
Stable Diffusion vs the alternatives
- Flux: from a lab founded by Stable Diffusion's original researchers; newer open models with better text and editing, some free for commercial use.
- Qwen: Alibaba's Qwen-Image models all scored well above SD 3.5 on Arena in Nov 2026.
- GPT Image and Nano Banana: far higher quality with no setup, but closed and online only.
- Midjourney: a polished subscription app, no local use.
- Krea and Leonardo AI: web apps for people who want custom styles without managing a local install.
Frequently asked questions
Is Stable Diffusion free? Yes. Stable Diffusion's weights are free to download and run on your own computer, and SD 3.5 is free even for commercial use if your annual revenue is under $1 million; Stability AI's hosted API charges per image.
What is the latest Stable Diffusion version? Stable Diffusion 3.5 (Large, Large Turbo and Medium), released in Oct 2024, is still the newest Stable Diffusion image model as of Nov 6, 2026.
Can I run Stable Diffusion on my PC? Yes. Stable Diffusion runs locally through free apps such as ComfyUI or the AUTOMATIC1111 web UI; SD 1.5 and SDXL work on modest graphics cards, and SD 3.5 Medium needs about 9.9 GB of VRAM.
What is the difference between SDXL and SD 3.5? SDXL (2023) is a 3.5 billion parameter model with years of community fine-tunes and add-ons; SD 3.5 (2024) is newer, up to 8.1 billion parameters, with better prompt following, under Stability's Community License.
Did Getty Images win against Stability AI? Mostly no. The UK High Court dismissed Getty's secondary copyright claim on Nov 4, 2025 and upheld only limited, historic trademark claims about watermarks generated by early Stable Diffusion versions.
Sources
- Introducing Stable Diffusion 3.5, Stability AI, Oct 22, 2024
- Announcing SDXL 1.0, Stability AI, Jul 26, 2023
- Stable Diffusion Public Release, Stability AI, Aug 22, 2022
- License, Stability AI, read Nov 6, 2026
- Pricing, Stability AI, read Nov 6, 2026
- News, Stability AI, read Nov 6, 2026
- Getty Images v Stability AI and judgment, High Court of England and Wales, Nov 4, 2025
- ComfyUI and Stable Diffusion web UI, GitHub, read Nov 6, 2026
- Text-to-Image leaderboard, Arena, read Nov 6, 2026
- Stable Diffusion explained (in less than 10 minutes), Render Realm, Mar 29, 2024 (video)
Moments
Stable Diffusion explained (in less than 10 minutes)
Render Realm
Related
-
Adobe FireflyAdobe Firefly is Adobe's AI app for images, video and audio, with its own Firefly models (trained, Adobe says, on licensed and public domain content) and partner models from Google, OpenAI and others. It has free daily generations, and paid plans start at $9.99 a month.
-
- ImagenImagen is Google's older line of text-to-image models, last Imagen 4 (2025). Google now calls it legacy: the Gemini API shut its Imagen models down from Aug 17, 2026 and points developers to Nano Banana instead.
-
KreaKrea is an AI creative app that generates images in real time as you type or draw, runs its own Krea 3 editing model and many partner image and video models, and upscales images. It is free with 100 compute units a day, and paid plans cost $9 to $200 a month.
-
Leonardo AILeonardo AI (Leonardo.Ai) is a Canva-owned creative app for AI images and video that offers its own Lucid models plus 80 or more partner models such as Nano Banana, GPT Image and Flux. It is free with 150 tokens a day, and paid plans cost $12 to $60 a month.
-
- SeedreamSeedream is ByteDance's family of image generation and editing models, made by its Seed team. The newest, Seedream 5.0 Pro, adds reasoning, infographics and layer separation, and costs $0.045 an image through BytePlus ModelArk.
-
See also
-
-
QwenQwen is Alibaba's family of AI models and its free chat app at chat.qwen.ai. The newest generation, Qwen3.8 (Aug 2026), includes the open-weight Qwen3.8-Max (2.4 trillion parameters) and an Apache 2.0 Qwen3.8-27B that runs on one large GPU.
-
-
Nano BananaNano Banana is Google's name for its Gemini image generation and editing models. The newest, Nano Banana 2.1, came out on Oct 6, 2026; it is free to use with limits in the Gemini app and costs about $0.034 per 1K image in the Gemini API.
-
MidjourneyMidjourney is a paid AI image and video generator from the independent lab Midjourney, Inc. Plans run from $10 to $120 a month with no free tier, and its default model since Jul 24, 2026 is V8.2.
-
Run AI LocallyTo run AI on your own computer, install Ollama or LM Studio (both free) and download an open model sized to your memory, such as Gemma 4, Qwen 3.5 or 3.8, or gpt-oss 20b. With 8 to 16 GB a small model runs well; a 24 GB graphics card or a 32 GB Mac runs 27B to 31B models.