Black Forest Labs Launches FLUX 3 and FLUX-mimic Robotics Model
Black Forest Labs has announced FLUX 3, a suite of multimodal flow models, alongside FLUX-mimic, a video-action robotics model developed in collaboration with mimic robotics. The company claims these models outperform competitors like ByteDance's Seedance 2.0, Gemini Omni, and Grok Imagine. This release bridges generative AI and physical robotics, demonstrating how advanced video generation models can serve as world models for real-world automation. The deployment of FLUX-mimic on Audi's production lines highlights the immediate industrial viability of these multimodal models. FLUX 3 utilizes a "Self-Flow" approach to align multimodal generation and understanding within a single architecture. FLUX-mimic leverages this backbone to perform complex tasks like dexterous manipulation and handling of soft materials.
## BACKGROUND
Black Forest Labs is known for its state-of-the-art image and video generation models, previously releasing the FLUX.1 series. Flow-based models and video-action models represent a shift towards training AI to understand physical dynamics, allowing robots to plan actions by predicting video frames of their environment.