The FLUX.2 [klein] model series is our fastest image model to date. FLUX.2 [klein] unifies image generation and editing capabilities within a single compact architecture, delivering industry-leading image quality with end-to-end inference times of under a second. Built specifically for applications that require real-time image generation without sacrificing quality.

FLUX.2 [klein] 9B is a 9-billion-parameter rectified flow Transformer model capable of generating images from text descriptions and supporting multi-reference image editing.

This is our flagship small model, defining the Pareto frontier between quality and latency across text-to-image generation, single-reference image editing, and multi-reference image generation tasks. Its performance matches or even surpasses models up to 5 times its size—with an inference time of less than half a second. Based on a 9-billion-parameter flow model and utilizing an 8-billion-parameter Qwen3 text embedder, it is compressed via step-distillation to require just 4 inference steps.