black-forest-labs/flux-3-action-base
Robotics • Updated • 278 • 128
Open weights 7B world action model: base and shared encoders, SO-101 and DROID policies. Code: https://github.com/black-forest-labs/flux-action
Note Action adaptation base and shared frozen video/text encoders.
Note SO-101 LeRobot policy with saved processors and task-LoRA recipe.
Note Validated DROID policy at root; opt-in optimization variants in subfolders.