Model · Open weights

Stable Diffusion

Stability AI · Stable Diffusion 3.5 / SDXL family · Stability AI Community License (model-specific terms)

Stable Diffusion is Stability AI's family of open-weight image and video generation models. It suits creators and developers who need local generation, fine-tuning, or hosted image workflows.

Stability AI offers Stable Diffusion through downloadable weights on Hugging Face and through its hosted API and partner platforms.

Official site ↗ · Weights ↗

Reported scores

Linked and dated. Made With Models has not independently tested these numbers; vendor-run results are marked in the note.

Benchmark	Score	Measured by	As of
Artificial Analysis Image Arena Elo ↗	759±10	Artificial AnalysisModel-specific result for Stable Diffusion 3.5 Medium.	2026-10

Made with this model

Made With Models illustration of a vintage telephone handset floating in a dim hallway as its cord dissolves into photographic fragments; this is not a frame from the film.

The message you never answered, rendered as a dream

Paul Trillo’s short film about longing and regret: Stable Diffusion storyboards, more than 400 Gen-2 clips, and a voice-over rewritten as the images found their shape.

By @paultrillo

How it was made

Stable Diffusion
Runway Gen-2

Made With Models illustration of a 1980s CRT television floating in a dark field, its antenna bent like a tuning fork, leaves, wings and spores resolving out of static on the phosphor screen; this is not a photo of the installation.

Tuning a 1987 CRT into the denoising steps of an ecology

Sihwa Park rebuilt a 1987 GoldStar television so its antenna controls the denoising stages of generated ecological scenes: move the antenna and noise resolves into extinct, endangered and speculative organisms across Past, Present and Future channels.

By Sihwa Park

How it was made

SDXL image sequences
Stable Audio Open
1987 GoldStar CRT
Raspberry Pi