German artificial intelligence lab Black Forest Labs has launched its third-generation flagship model, Flux 3, marking its first move beyond static visuals into video generation and physical-world manipulation. The model can generate synchronized audio-video clips up to 20 seconds long from text prompts. Through a partnership with Zurich-based robotics company Mimic Robotics AG, it has spawned an action model called Flux-mimic that directly commands robotic arms to perform delicate assembly tasks. Flux 3 was trained on image, video, and audio data within a single architecture. In early blind comparison tests, human evaluators preferred Flux 3's output over Runway Gen-4.5 in 77% of direct comparisons; against Luma Ray 3.2, that preference ratio soared to 93%. In matchups against Gemini Omni and Seedance, Flux 3 held a slight edge with a 52% win rate. Automaker Audi AG has become one of the first industrial testers of this technology, with the two companies validating the robot's ability to handle complex tasks such as installing flexible door seals.