Fetching from the wire…
01
02
03
04
05
06
07
08
Models2026-07-24 · source-backed
Announced alongside FLUX 3, it bolts a lightweight action decoder onto intermediate FLUX 3 features to make a video-action model, beating prior VLA models even with the backbone frozen. Deployed at Audi for kitting, component insertion and flexible-material manipulation, with failure recovery it was never explicitly shown. Generative video pretraining is being harvested as a robotics world model rather than trained separately. No public release date.
Each link below shares sources, entities, or timing with this story.
Shared entity: FLUX / Same source domain / Shared topic
Both cover FLUX; reported by the same outlet (bfl.ai); overlapping topics (action, backbone, flux).
Shared entity: VLA / Shared topic / Earlier coverage / Tension
Both cover VLA; overlapping topics (beating, model); earlier VLA coverage from 2026-07-20.
Shared entity: Flux / Shared topic / Earlier coverage
Both cover Flux; overlapping topics (component, flux); earlier Flux coverage from 2026-06-08.
Shared entity: VLA / Earlier coverage
Both cover VLA; earlier VLA coverage from 2026-03-05.
Both cover VLA; earlier VLA coverage from 2026-03-05.
Shared topic / Tension
Overlapping topics (action, model); pushes against this story (but).
Overlapping topics (action, model); pushes against this story (against).
Overlapping topics (backbone, model); pushes against this story (against).