DEPLOYDatabase

Brain

FLUX 3 Action

FLUX 3 Action is Black Forest Labs' open-weights 7B World Action Model for robotics and related visual control tasks. Per company claims (Sep 23 2026), it takes camera frames, robot state, and language instructions, jointly predicts future frames and action chunks, and is positioned for fine-tuning on robot-specific demonstrations (including LeRobot / Jetson paths). Benchmark figures are company-claimed; public weight availability and license terms should be confirmed against BFL's release materials.

World model · Maturity: Research · Open source

Machine-readable surfaces

Architecture

7B World Action Model; joint future-frame + action prediction on FLUX 3 multimodal pretraining; smaller deployment-oriented architecture vs full FLUX 3 video stack

Key facts

FLUX 3 Action (Sep 2026, company-claimed)
Company (@bfl_ai Sep 23 2026): FLUX 3 Action is an open-weights 7B World Action Model that predicts robot actions and future scenes from camera observations plus instructions. Company claims first place on RoboLab vs prior open best (+6.1 percentage points), 56% fewer parameters, and up to 3.95x faster; fine-tuning recipes and LeRobot / Jetson integration with NVIDIA and Hugging Face. VentureBeat (same day) relays company-supplied RoboLab-120 overall success ~42.92% and notes NVIDIA public leaderboard had not yet listed the result at press time. Official bfl.ai FLUX 3 page still describes Action rollout via selected partners beginning with mimic robotics; treat benchmark and open-weight timing as company-claimed pending leaderboard confirmation and public weight drop
FLUX 3 multimodal scope (primary Black Forest Labs)
Black Forest Labs' FLUX 3 page describes one multimodal foundation model for generating and understanding video, audio, images, and actions in a unified architecture. FAQ text says FLUX 3 Action is rolling out through selected research and commercial partners beginning with mimic robotics, separate from FLUX 3 Video API access ([Black Forest Labs](https://bfl.ai/models/flux-3))
FLUX 3 Video generation limits (primary Black Forest Labs)
The same FAQ says FLUX 3 can generate clips up to 20 seconds in a single generation across HD through UHD/4K bands, with optional native audio including multilingual dialogue. Draft mode is described as a faster lower-cost HD preview. Treat limits and pricing as BFL-stated product terms ([Black Forest Labs](https://bfl.ai/models/flux-3))
FLUX 3 Action-Prediction modality (primary BFL)
Black Forest Labs' FLUX 3 model page presents FLUX 3 as one multimodal model covering Image, Video, Audio, and Action-Prediction. The page states commercial partners beginning with mimic robotics, and that open-weight access to the FLUX 3 Dev multimodal backbone is coming soon. Treat Action-Prediction robotics positioning and partner naming as company-stated marketing on the product page ([Black Forest Labs](https://bfl.ai/models/flux-3))

Developed by (1)

Common questions

What is FLUX 3 Action?
FLUX 3 Action is Black Forest Labs' open-weights 7B World Action Model for robotics and related visual control tasks. Per company claims (Sep 23 2026), it takes camera frames, robot state, and language instructions, jointly predicts future frames and action chunks, and is positioned for fine-tuning on robot-specific demonstrations (including LeRobot / Jetson paths). Benchmark figures are company-claimed; public weight availability and license terms should be confirmed against BFL's release materials.
Who developed FLUX 3 Action?
FLUX 3 Action is credited to Black Forest Labs on the DEPLOY registry. Each developer attribution is verified via primary sources.
Is FLUX 3 Action open source?
Yes. FLUX 3 Action is recorded as open-source / open-weights on the DEPLOY registry, meaning model weights or source code are publicly available.
What type of AI is FLUX 3 Action?
FLUX 3 Action is a world model, built on a 7B World Action Model; joint future-frame + action prediction on FLUX 3 multimodal pretraining; smaller deployment-oriented architecture vs full FLUX 3 video stack architecture on the DEPLOY registry.
What is FLUX 3 Action's maturity stage?
FLUX 3 Action is at the research stage on the DEPLOY maturity ladder. Research stage means active development without commercial deployments on file.
Which robots run on FLUX 3 Action?
No robot models on the DEPLOY registry are recorded as running FLUX 3 Action. DEPLOY wires brain-to-model connections only when the wiring is verifiable from primary sources; absence may reflect pre-deployment or unverified manufacturer claims.

Sources (3)

  1. https://x.com/bfl_ai/status/2102816874782241174 · 2026-09-23
  2. https://bfl.ai/models/flux-3 · 2026-09-23
  3. https://venturebeat.com/ai/black-forest-labs-debuts-flux-3-action-an-open-weights-ai-robotics-model-that-tops-the-leaderboard-at-half-the-size-of-its-competition · 2026-09-23
Methodology: Verified · 3 sources (1 primary) · last reviewed 2026-09-29

Verification posture

Verified

High confidence

Review state

Stable

Last reviewed 2026-09-29

Maturity + lifecycle

Maturity stage: research

Sources by quality tier

1
unclassified
Unclassified source
1
primary-company-ir
Company IR disclosure
1
secondary-industry-publication
Industry publication

The framework is documented at /methodology. Corrections at /corrections. Reviewer: DEPLOY editorial team.

Methodology surface for FLUX 3 Action.

Canonical ID 93a6c093-d87a-4149-ade7-9dec2ad2e237