Black Forest Labs Launches FLUX.3 Action, an Open Robotics AI Model
Black Forest Labs has released FLUX.3 Action, an open-source AI model designed for robotics control. The model processes visual and text inputs to generate real-time action commands. The launch targets researchers and developers building autonomous systems who need a transparent, modifiable baseline.
What FLUX.3 Action Does
FLUX.3 Action takes a camera feed or image plus a text instruction — for example, “pick up the red cube” — and outputs precise motor controls or joint angles. It operates on a diffusion-based architecture similar to the company’s image-generation models, but adapted for sequential decision-making.
The model is fully open-weight and available under a permissive license. Developers can fine-tune it on their own robot hardware without vendor lock-in.
Key capability: The model can generalize to new objects and environments without retraining, using only a few demonstration examples.
Why Open Robotics AI Matters
Proprietary robotics models force developers to rely on black-box APIs or expensive licensing. An open model allows:
- Full transparency — engineers inspect and modify every weight and training step.
- Local deployment — no internet dependency for inference, reducing latency and privacy risks.
- Community-driven improvement — any team can contribute datasets, fine-tunes, or safety checks.
Black Forest Labs positions FLUX.3 Action as a direct competitor to closed-source systems from major robotics firms, aiming to accelerate research in manipulation, navigation, and human-robot interaction.
Technical Details
The model was trained on a large dataset of simulated and real robot trajectories. It uses a transformer backbone with a diffusion head that iteratively refines action predictions. Key specs include:
- Input modalities: RGB images (single or multi-view), natural language commands
- Output format: Continuous action vectors (e.g., end-effector position, gripper force)
- Inference speed: Sub-100ms on a consumer GPU for simple tasks
- Supported hardware: Franka, UR5, and other common research arms (with adapter scripts)
Licensing and Availability
FLUX.3 Action is released under an Apache 2.0 license. Weights, inference code, and fine-tuning scripts are available on GitHub and Hugging Face. The company also provides a training pipeline that runs on a single A100 GPU.
Warning for production use: The model has not been safety-certified for real-world industrial applications. Black Forest Labs recommends thorough testing in simulation before robotic deployment.
Background
Black Forest Labs previously released the FLUX series of image-generation models, which gained attention for their high quality and open-source ethos. FLUX.3 Action extends that approach into embodied AI — a field where most models remain proprietary.
The company plans to release benchmark results and a leaderboard for open robotics models in the coming months.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.