FLUX.1-Turbo-Alpha is an 8-step distilled Lora based on the FLUX.1-dev model, released by the Alimama Creative Team. This model employs a multi-head discriminator to enhance distillation quality and can be used for text-to-image (T2I), denoising control networks, and other FLUX-related models. It is recommended to set the guidance scale to 3.5 and the Lora scale to 1. The model is trained on 1M open source and internal images, using adversarial training to improve quality, with the original FLUX.1-dev transformer fixed as the backbone of the discriminator and adding multi-heads at each layer.