AI models and robotics
Black Forest Labs unveils multimodal FLUX 3
Black Forest Labs introduced FLUX 3, a unified model trained across image, video and audio, as well as FLUX-mimic, a video-action model developed with Mimic Robotics and tested in manufacturing environments including Audi.
AS1 News
Black Forest Labs introduced FLUX 3 on July 23, 2026, expanding its FLUX family beyond image generation. The company described FLUX 3 as a unified model trained across image, video and audio.
The announcement also included FLUX-mimic, a video-action model developed with Mimic Robotics. According to the supplied event package, FLUX-mimic has been tested in manufacturing environments, including Audi.
The release matters because it moves a major visual-model family toward native audiovisual generation, world modeling and robotic action prediction. Pairing a multimodal foundation model with a system focused on predicting actions from video also connects generative AI research more directly with physical automation.
The introduction of FLUX 3 and FLUX-mimic, their stated training modalities, the collaboration with Mimic Robotics and testing in manufacturing settings including Audi are confirmed in the event package. The package does not establish comparative performance, benchmark results, commercial availability, deployment plans or the extent of testing at Audi. Those points therefore remain uncertain.
FLUX 3 broadens the FLUX family from image generation toward unified image, video and audio capabilities, while FLUX-mimic extends the initiative into world modeling and robotic action prediction. No commercial or financial impact is yet confirmed.