Robbyant open-sources AI model designed to control many robot types
Robbyant’s LingBot-VLA 2.0 release targets a central robotics problem: building intelligence that can generalise across different robot bodies and hardware platforms.
Robotics is still highly fragmented. Different machines have different sensors, kinematics and control systems, making intelligence difficult to reuse.
What happened
Robbyant released and open-sourced LingBot-VLA 2.0, a vision-language-action model designed to generalise across multiple robot types.
The company says the model was pretrained on 60,000 hours of physical data gathered from 20 robot morphologies across 17 manufacturers.
It supports multiple forms including single-arm, dual-arm, bipedal and wheeled systems and is being piloted in retail sorting, logistics and industrial automation.
Why it matters
A major robotics bottleneck is that intelligence is often built separately for each hardware platform.
A model that can generalise across multiple robot bodies could reduce duplication and make physical-AI software more reusable.
The open-source release matters too because it could support a broader ecosystem around shared embodied-AI models rather than only closed, vertically integrated robots.
The bigger picture
Robotics is beginning to search for its equivalent of a foundation-model layer.
The long-term opportunity is a reusable intelligence stack that can transfer across machines, tasks and environments. Robbyant’s release is a direct bet on that architecture.