tl;dr: Black Forest Labs released FLUX 3 in early access, a multimodal foundation model that jointly generates and understands video, audio, and images within a unified architecture based on their "Self-Flow" approach. It can produce up to 20-second videos with native audio from text, image, or video inputs, and early evaluations claim win rates over competitors like Kling v3 Pro (60%), Runway Gen-4.5 (77%), and Luma Ray 3.2 (93%). The same backbone is also being used for robotics action prediction via a partnership with mimic robotics, with open-weight "FLUX 3 Dev" access planned.