Black Forest Labs Launches FLUX 3 Action, a Multimodal Foundation Model
Black Forest Labs has made significant strides in the physical AI space with its latest innovation, FLUX 3 Action. This multimodal foundation model boasts an impressive 7-billion-parameter open-weight robotic model that outperforms larger competitors on the RoboLab-120 benchmark.
The model's capabilities are built upon the company's newly unveiled FLUX 3 foundation model, which integrates images, video, and audio using a technique called Self-Flow. This architecture pairs a video-prediction backbone with a streamlined decoder that translates visual sequences into robotic control commands.
FLUX 3 Action has already been tested on soft-body manipulation tasks at Audi, where it achieved a 95% success rate. Additionally, Black Forest Labs has initiated a partnership with mimic robotics to develop FLUX-mimic, the first commercial deployment channel for the action capabilities.