Stable Diffusion Statistics: Downloads, Usage & Ecosystem
Stable Diffusion: ~80% of all AI-generated imagery, 12.59B+ cumulative images (2023 baseline), 50M+ total downloads, 500K+ models/LoRAs on Civitai, ~2B images generated monthly across SD-based platforms.
Stable Diffusion is the open-source workhorse of AI image generation. While paid tools win brand-preference surveys, the Stable Diffusion ecosystem produces the bulk of actual output - roughly 80% of all AI-generated imagery runs on SD-derived models through self-hosted installs, third-party apps, and APIs. Its open weights spawned the largest model-sharing community in generative AI.
Key statistics
Stable Diffusion and its derivatives account for approximately 80% of all AI-generated images worldwide, driven by self-hosted installs and the thousands of apps built on its open weights.
Everypixel Journal attributed ~12.59 billion cumulative images to the Stable Diffusion ecosystem in its 2023 baseline analysis - a count that has expanded substantially since.
Stable Diffusion models have been downloaded over 50 million times across all platforms and versions, including ~28M on Hugging Face for SDXL and SD 1.5 combined.
Civitai hosts over 500,000 fine-tuned models and LoRAs built on Stable Diffusion, with ~214 million cumulative model downloads on the platform.
An estimated 2+ billion images are generated monthly across all Stable Diffusion-based platforms and local installations - roughly 70 million per day.
Over 2,500 unique applications have been built on Stability AI’s APIs, in addition to the uncountable self-hosted deployments.
How the Stable Diffusion ecosystem layers compare
Stability AI’s official channels are the smallest layer of the ecosystem: roughly 2 million images per day versus about 70 million per day across the full open-source pipeline. Most usage happens in third-party tools and self-hosted installs that no brand survey captures.
| Ecosystem layer | Scale | What it measures | Source |
|---|---|---|---|
| Official channels (DreamStudio, API) | ~2M images/day | Stability AI’s own traffic | SQ Magazine / Everypixel |
| Full ecosystem output | ~2B images/month (~70M/day) | All SD-based platforms and local installs | Apatero |
| ComfyUI | 4M+ users | Largest node-based SD interface | Apatero |
| Automatic1111 | ~3M users | Most popular web UI for local SD | Apatero |
| Civitai community models | 500K+ models, ~214M downloads | Fine-tunes and LoRAs | Civitai platform data |
| Base model downloads | 50M+ (28M on Hugging Face) | SDXL + SD 1.5 official weights | Apatero |
Layer comparison compiled June 2026. Open-source usage cannot be measured directly; ecosystem-wide figures are analyst estimates, while Civitai and Hugging Face counts are platform-reported.
How large is the Stable Diffusion ecosystem?
The ecosystem extends far beyond Stability AI’s own channels: official channels produce ~2 million images daily, while the broader open-source pipeline produces more than 13 times Midjourney’s output.
Stability AI’s own channels (DreamStudio and the Stable Diffusion API) generate roughly 2 million images daily - a small fraction of total ecosystem output.
Analysts estimate 15–20 million monthly active users across SD-based tools, with ComfyUI alone reaching 4+ million users and Automatic1111 around 3 million.
SDXL became the dominant version, accounting for about 65% of new Stable Diffusion generations before the SD 3.5 series (October 2024) introduced the MMDiT architecture.
Stability AI raised approximately $225 million through 2024 to fund open-weight model development.
What did the Getty Images lawsuit decide about Stable Diffusion?
The UK High Court handed down the first British ruling on generative AI training in Getty Images v Stability AI on November 4, 2025 - and largely rejected Getty’s copyright claims.
The High Court of England and Wales rejected Getty’s core copyright claim, reasoning that model weights do not store copies of training images. It found only limited trademark infringement on older SD v1.x/v2.x outputs displaying Getty watermarks.
The Getty case is one of more than 70 copyright suits filed against AI companies through 2025 across text, image, audio, and video models.
Stable Diffusion shows only 15.1% in user-preference surveys despite producing most actual output - open-source distribution is largely invisible to brand tracking.
How we compiled this data
We separated three layers that most coverage conflates: Stability AI the company, the official Stable Diffusion model releases, and the open-source ecosystem built on those weights. Figures were cross-referenced across Everypixel Journal’s image-count baseline, Apatero’s 2025 usage analysis, Civitai’s public platform counters, and SQ Magazine’s survey data. For the Getty v Stability AI ruling we cite the UK High Court judgment directly rather than press summaries. Last full review: June 12, 2026.
Before you cite these numbers
- Open-source usage is structurally unmeasurable. Anyone can run Stable Diffusion locally with no telemetry, so the ~80% volume share and ~2B/month output are analyst estimates, not measured counts.
- The 12.59B cumulative image figure is Everypixel Journal’s August 2023 baseline. The true 2026 cumulative count is much higher but no rigorous updated count exists.
- Stability AI’s finances do not describe the ecosystem. The company raised ~$225M and went through well-documented financial turbulence in 2024, while the open ecosystem kept growing independently.
- Civitai’s 500K+ model count includes merges and minor LoRA variants, so it overstates the number of meaningfully distinct models.
Frequently asked questions
How many images has Stable Diffusion generated?
Everypixel Journal attributed approximately 12.59 billion cumulative images to the Stable Diffusion ecosystem in its 2023 baseline analysis. With an estimated 2+ billion images generated monthly across SD-based platforms since, the cumulative total is now far higher.
Is Stable Diffusion still the most-used AI image model?
By output volume, yes - roughly 80% of AI-generated imagery runs on Stable Diffusion or its derivatives. By brand preference, Midjourney (26.8%) and DALL·E (24.4%) score higher because most SD usage happens inside third-party apps and self-hosted installs.
How many times has Stable Diffusion been downloaded?
Over 50 million times across all platforms and versions, including roughly 28 million downloads on Hugging Face for SDXL and SD 1.5 combined. Community models on Civitai add another ~214 million downloads.
Did Getty Images win its lawsuit against Stability AI?
Mostly no. In its November 4, 2025 judgment, the UK High Court rejected Getty’s core copyright claim because model weights do not store the underlying images. The court found only limited trademark infringement on older Stable Diffusion v1.x and v2.x outputs that reproduced Getty watermarks.
Is Stable Diffusion still relevant in 2026?
Yes, as infrastructure more than as a brand. Newer models (FLUX, GPT Image, Nano Banana) lead benchmarks, but the Stable Diffusion ecosystem still runs an estimated ~2 billion images monthly through ComfyUI, Automatic1111, and 500K+ community fine-tunes - and remains the default base for self-hosted, customized pipelines.
Sources
Figures on this page are compiled from the following publishers and reports. Where sources disagree, we present the range and note the methodology difference.