Born from foundational research, BFL continually build state-of-the-art technology that advances how the world is seen and understood.
This is the 3rd launch from FLUX by Black Forest Labs. View more
FLUX 3 Action
Launched this week
Gemini 3.8 Flash TTS and Flash-Lite TTS turn text into expressive speech. Create custom voices from natural-language prompts, control pacing, emotion and dialect, and generate long-form audio or two-speaker conversations.
Launch Team
Dropbox SignSimple, secure eSignatures for the way your team works.
Promoted
FLUX 3 Action is Black Forest Labs’ 7B World Action Model for robot control. It jointly predicts what the robot should do and what the world will look like after those actions.
Give it a natural-language instruction, camera images, and joint positions. It predicts motor commands and their visual outcomes, the robot executes them, then the model looks again and plans from the fresh observation.
What you get:
7B open-weight model
Joint video and action prediction
Natural-language task instructions
Camera and robot-state inputs
Predictions of future visual outcomes and robot actions
Control loop that can re-plan from fresh observations
Support for manipulation tasks and multiple robot embodiments
Single-step and guidance-distilled checkpoints
FP8 support for inference efficiency
Try FLUX 3 Action model