FLUX.1 [schnell]
FLUX.1 Schnell is a 12B text-to-image model trained with latent adversarial diffusion distillation. It generates an image in one to four sampling steps.
Model details
Schnell is released under Apache 2.0. Its short sampling recipe differs from the guidance-distilled Dev model. It can be run locally through the BFL reference implementation, Diffusers or ComfyUI.FLUX.1 [schnell] official documentation
Practical use
Practical guidance. Schnell is useful for fast local drafts and workflows that need Apache-licensed weights. Check the quality of small details and text after selecting a prompt. A short sampling run can be efficient even when a separate model is used for the final image.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | black-forest-labs/FLUX.1-schnell |
| Weight license | Apache 2.0FLUX.1 [schnell] official documentation |
Related models
- FLUX.2 [klein] 4B
Black Forest Labs · FLUX.2
Four-billion-parameter image generation and reference-editing model distilled to four sampling steps.
Launch date not confirmedAPI · Downloadable · Provider product - FLUX.1 [dev]
Black Forest Labs · FLUX.1
Twelve-billion-parameter text-to-image transformer trained with guidance distillation.
Launch date not confirmedAPI · Downloadable