FLUX 3 Action SO-101
FLUX 3 Action SO-101 is a 7B policy for the SO-101 robot setup. It uses scene and wrist cameras, robot state and a task instruction to predict action sequences together with future frames.
Model details
The policy has six state/action dimensions and uses joint deltas with an absolute gripper command. It predicts 42 actions, executes 32 at 30 Hz and replans. Saved calibration and normalization files are part of the checkpoint contract. The card recommends task-specific adaptation.FLUX 3 Action SO-101 official documentation
Practical use
Practical guidance. Match the camera order, joint units and saved normalization before evaluating this policy. Record demonstrations for the actual arm and task rather than assuming zero-shot transfer to a new setup. Keep physical limits in the robot controller; the model does not enforce them.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | black-forest-labs/flux-3-action-so101 |
| Weight license | FLUX Kommunity License 1.0; text encoder separately under Apache 2.0.FLUX 3 Action SO-101 official documentation |
| Cameras | Scene, then wrist.FLUX 3 Action SO-101 official documentation |
| Access | Downloadable policy; accepts the model conditions on Hugging Face.FLUX 3 Action SO-101 official documentation |
Related models
- FLUX 3 Action Base
Black Forest Labs · FLUX 3
Seven-billion-parameter world-action model that jointly predicts future visual frames and robot action sequences.
Launch date not confirmedDownloadable - FLUX 3 Action DROID
Black Forest Labs · FLUX 3
Seven-billion-parameter robot policy that generates joint commands and future frames from DROID-style observations.
Launch date not confirmedDownloadable