DownloadFLUX.1-schnell is a 12 billion parameter text-to-image model developed by Black Forest Labs, designed for high-quality image generation using rectified flow transformers. It produces competitive visual results with strong prompt adherence, rivaling closed-source models in just 1 to 4 inference steps. Trained using latent adversarial diffusion distillation, the model is optimized for both quality and speed. It is released under the Apache 2.0 license, allowing commercial, scientific, and personal use. The model can be accessed via the FluxPipeline in Hugging Face’s diffusers library and is compatible with local workflows like ComfyUI. Available through several inference providers including Replicate and fal.ai, FLUX.1-schnell is well-documented and supported for developers. While capable, the model is subject to limitations such as occasional prompt misalignment and potential amplification of societal biases, and it may not be used in harmful, exploitative, or deceptive applications.
Features
- 12B-parameter rectified flow transformer for text-to-image generation
- Generates high-quality images in just 1–4 inference steps
- Trained with latent adversarial diffusion distillation
- Compatible with Hugging Face diffusers and ComfyUI workflows
- Supports commercial use under the Apache 2.0 license
- API access via providers like Replicate, fal.ai, and bfl.ml
- Competitive prompt following with fast inference
- Out-of-the-box reference implementation and sampling code provided
Project Samples
