Gaode Fully Opens Source Embodied Operation Base Model ABot-M0, Adapts to Multiple Robot Forms

In daily scenarios, robots come in various forms (like industrial robotic arms and home service robots), and a universal brain adapting to different forms has long been an industry challenge. Recently, Gaode announced the full open-source of the embodied operation base model ABot-M0, the world's first robot embodied operation base model based on a unified architecture, enabling a "universal brain" to adapt to multiple robot forms.

The model performs excellently in authoritative tests like Libero, with an 80.5% success rate on the Libero-Plus benchmark, nearly 30% higher than Pi0. This open-source covers three dimensions: data (UniACT dataset), algorithms (AML algorithm, etc.), and models (pre-trained models and toolchains). Developers can quickly adapt to industrial and home scenarios without building a framework from scratch. Gaode's tech lead said they hope the model becomes a bridge between academia and industry, giving every robot a universal brain.