Stable Diffusion 3.5 (Large, Large Turbo, Medium)
Stability AI · Image model
Still the newest Stable Diffusion image model as of October 2026; Stability AI has announced no Stable Diffusion 4. It remains popular for local and customized workflows because the weights are open. Stability's 2026 releases have focused on audio (Stable…
Legacy model Open weights
ReleasedOct 22, 2024
Context window—
Input → outputtext → image
API price—
Consumer accessFree weights on Hugging Face under the Stability Community License (free for organizations/individuals under $1M annual revenue; Enterprise License above that).
Best for
Self-hosted, fine-tunable image generation on your own hardware.
Overview
Still the newest Stable Diffusion image model as of October 2026; Stability AI has announced no Stable Diffusion 4. It remains popular for local and customized workflows because the weights are open. Stability's 2026 releases have focused on audio (Stable Audio 3.0) and enterprise tools instead.
Strengths and limits
- Open weights; Large is 8B parameters
- Large Turbo generates in 4 steps; Medium runs on consumer GPUs (about 9.9 GB VRAM)
- Large ecosystem of fine-tunes, tools and optimized builds (TensorRT, ONNX, NIM)
- Quality well behind 2026 frontier models; not on the current Artificial Analysis top board
- Commercial use above $1M revenue needs an Enterprise License