Flux Fast is a speed-optimized text-to-image model from Pruna AI, based on Black Forest Labs’ FLUX.1-dev architecture. It applies Pruna AI’s compression, caching, and compilation techniques to significantly accelerate inference while maintaining high-quality image generation. The model is designed for applications where low generation latency and high-volume image creation are important. Replicate currently reports example inference times around 0.8–1.4 seconds, although actual latency can vary depending on settings and infrastructure.