FLUX 3 Action An open-weight 7B world-action model derived from Black Forest Labs’ multimodal FLUX 3 backbone. It jointly predicts chunks of robot actions (about 2 seconds at 15 Hz) and future video frames from camera images, robot state, and text instructions, enabling physical AI control for arms, drones, and games. It sets a new state-of-the-art on the RoboLab benchmark with fewer parameters and up to 3.95x faster runtime than prior open models.