Black Forest Labs releases FLUX 3 Action, a 7B open-weight model for robot control

FLUX 3 Action is a 7-billion-parameter model that takes camera images, a robot current state, and a natural-language instruction, and predicts a chunk of actions alongside future video frames.

an open-weight 7B world-action-model ... Given an instruction such as put the red cube in the left bin, the model uses camera images and joint positions to predict motor commands and their visual outcomes together.

Black Forest Labs reports 42.92 percent on RoboLab-120, 6.1 points above the previous best open model, NVIDIA 16-billion-parameter Cosmos3-Nano-Policy (36.8 percent), with less than half its parameters.

Table 1: RoboLab-120 overall success rates. FLUX 3 Action, not closed source, 42.92%, 7B ... OASIS WAM, closed source, 39.0% ... Cosmos3-Nano-Policy, not closed source, 36.8%, 16B ... at less than half the parameters of the previous best open model, while running up to 3.95x faster.

Black Forest Labs publishes the weights on Hugging Face as an open-weight model.

Download the weights (links to huggingface.co/collections/black-forest-labs/flux-3-action) ... an open-weight 7B world-action-model