Stable Diffusion 3.5 Medium
Image modelOpen weightsCard summarizedWith 2.5 billion parameters, the model delivers an optimal balance between prompt accuracy and image quality, making it an efficient choice for fast high-performance image generation.
Source: Stable Diffusion 3.5 Medium model card (Hugging Face), SD 3.5 launch post and API reference
Where to call it
- Stability AI Platform API
sd3.5-medium - Self-host (weights)
stabilityai/stable-diffusion-3.5-medium
Good for
- Image generation on standard consumer hardware
- Balance of prompt accuracy and image quality at low cost
- Artwork, design and creative or educational tools
Overview
SD 3.5 Medium
With 2.5 billion parameters, the model delivers an optimal balance between prompt accuracy and image quality, making it an efficient choice for fast high-performance image generation.
- Max resolution
- 1 MP via API
- Aspect ratios
- 9 options
- Modes
- text-to-image, image-to-image
- Released
- 29 Oct 2024
Stable Diffusion 3.5 Medium is Stability AI's mid-size open-weights image model, released on 29 Oct 2024. It is a 2.5-billion-parameter MMDiT-X model that generates 0.25 to 2 megapixel images and needs 9.9 GB of VRAM (excluding text encoders), so it runs on consumer GPUs. Stability's API also serves SD 3.5 Flash, a distilled 4-step version of Medium that has no open weights. launch post; model card; API reference
sd3.5-medium- Typical price
- $0.035 / image
- Max resolution
- 1 MP via API (default 1024×1024); weights generate 0.25–2 MP
- Aspect ratios
- 21:9, 16:9, 3:2, 5:4, 1:1, 4:5, 2:3, 9:16, 9:21
- Weights
- Open · Stability AI Community License (free under $1M annual revenue; Enterprise License above)
- Released
- 29 Oct 2024
- text-to-image
- image-to-image