BFL Releases FLUX 3 Action for Robot Control
BFL Releases FLUX 3 Action for Robot Control
Black Forest Labs introduced FLUX 3 Action, a 7B video model fine-tuned on robot recordings to predict both future frames and joint actions.
Black Forest Labs (BFL) has fine-tuned the FLUX 3 video model foundation on robot operation recordings, where camera frames are synchronized with joint positions and actions. The 7B model learns to jointly denoise future frames and movement sequences.
Given a text task, current frames, and the robot state, the model predicts joint and gripper movements together with their visual outcome. The robot executes part of the commands, receives new frames, and recomputes the next movements.
On the RoboLab-120 simulation, it took first place with 42.9% successful attempts, compared with 36.8% for the 16B Cosmos 3 Nano. Fine-tuning is required for a new robot. Weights are available under the FLUX Kommunity License, and the code is released under Apache 2.0.
Why it matters
AnalysisFLUX 3 Action shows how video models can become policy models for embodied AI, predicting visual consequences and robot actions in one loop. Open weights and Apache-licensed code make it easier for researchers to adapt the approach to new robots.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Google releases Gemini 4 Carbon
Google announced the Gemini 4 Carbon AI model and made it available to a wider audience, effectively limited to AI Ultra subscribers paying $250.
JSON — Structured Data for Programs, APIs, and AI Agents
A complete hands-on guide to JSON: syntax specifications, serialization via parse and stringify, REST API consumption, JSON Schema validation, and structured outputs for LLMs.
DOCX and XLSX Formats: When AI Needs to Deliver Production-Ready Documents
A comprehensive hands-on guide to generating and processing DOCX documents and XLSX workbooks with AI: comparing against Markdown and CSV, prompt engineering, formula design, and code automation.
DeepSeek (Chinese Breakthrough in Open AI)
A series of groundbreaking open models from the Chinese lab DeepSeek (V3, R1). Demonstrates top-tier coding and reasoning capabilities at a cost 10–20 times lower than Western commercial counterparts.