Black Forest Labs launched FLUX3, a multimodal foundation model in early access, unifying image, video, and audio learning via a self-supervised flow framework. It generates up to 20-second videos with native audio and supports text-to-video, image-to-video, and more. In benchmarks, FLUX3 achieved a 69% win rate against Grok Imagine Video, 52% against Seedance 2.0 and Gemini Omni Flash, and extends to robot behavior prediction.