Stability AI Intermediate

Stable Diffusion 3 Medium

Stable Diffusion 3 Medium is Stability AI's June 2024 MMDiT image checkpoint. Its original API is deprecated, but the open weights remain available.

Image GenerationTextImage Open Source, Paid (API)

In plain English

What is this model and why does it matter?

Stable Diffusion 3 Medium is Stability AI's June 2024 open-weight MMDiT image model. The original SD3.0 API is deprecated, but the checkpoint remains downloadable.

Historical SD3 workflowsLocal image generationCreative researchCompatibility testing

Model overview

Stable Diffusion 3 Medium: features, use cases and important details

Stable Diffusion 3 Medium is Stability AI’s June 2024 text-to-image checkpoint based on the Multimodal Diffusion Transformer architecture. The exact official repository is stabilityai/stable-diffusion-3-medium.

What the model improved

Stability positioned SD3 Medium as an improvement in image quality, typography, complex prompt understanding and resource efficiency. The model uses three pretrained text encoders and was released under the Stability AI Community License for local and self-hosted use.

API status

The original Stable Diffusion 3.0 APIs are no longer a current standalone service. Stability states that, as of April 17, 2025, SD3.0 API calls were deprecated and automatically rerouted to Stable Diffusion 3.5 APIs.

Who should still use it

The checkpoint remains relevant for reproducibility, existing SD3 workflows and local experimentation. New hosted Stability integrations should start with the SD3.5 family instead.

Get started

How to use this model

Accept the Stability Community License.
Download stabilityai/stable-diffusion-3-medium or its Diffusers-compatible weights.
Run locally with a supported image-generation pipeline.
Use SD3.5 for new Stability API integrations.
Review output quality and licensing requirements before commercial deployment.

Copy and try

Example prompts

A modern magazine cover with bold geometric typography and a studio portrait.
A cinematic architectural visualization of a timber library at dusk.
A surreal botanical illustration with precise layered detail.

Capabilities

What it can do

Text-to-image generation
MMDiT architecture
Typography improvements
Complex prompt understanding
Local deployment

Best for

Practical use cases

Creative research
Local image generation
Legacy SD3 projects
Design experimentation

Pricing

What does it cost?

The original Stable Diffusion 3.0 API was deprecated and rerouted to Stable Diffusion 3.5. Do not present historical SD3 Medium API pricing as current.

Simple summaryLocal use depends on GPU infrastructure and the Community License. The old SD3.0 API no longer has standalone current pricing.

What stands out

Official downloadable checkpoint
Improved typography over older Stable Diffusion generations
Community License

Things to consider

Original API deprecated
Superseded by SD3.5
Can require substantial GPU memory

Limitations

Important restrictions and trade-offs

Old API calls are rerouted to newer models
Image generation can still produce anatomy, text or composition errors
Commercial license conditions depend on organization revenue

SimplifyAITools verdict

Our editorial take

Keep as a verified historical SD3 checkpoint, but direct new API users to Stable Diffusion 3.5 rather than presenting SD3 Medium as a current hosted model.

References