Neura Robotics Unveils Neuron AI: A Multimodal Robotic Brain With Hearing, Touch, and Memory

Metzingen, Germany—a town once known mainly for luxury outlet stores—is now positioning itself at the center of a robotics revolution. Neura Robotics has officially unveiled Neuron AI, its flagship model, which the company describes without reservation as the world’s leading and “only” multimodal large model for physical AI. The launch underscores the company’s ambition to set the standard for embodied intelligence and to move beyond task-specific automation.

With $1.4 billion (about RMB 10 billion) in funding, Neura is betting on a single proposition: building a universal “ultimate brain” for robots of different shapes and sizes. David Reger, founder and CEO of Neura Robotics, said intelligent robots need a “brain” that can integrate all sensory inputs and remember what has happened, and that Neuron AI was created for precisely that purpose. The company argues that general-purpose intelligence, rather than narrow models, is the key to scaling robotics across diverse hardware.

At the technical level, Neuron AI is a perception-language-action (PLA) model. In plain terms, it does more than “see” the world—it “experiences” it. The model fuses 3D vision, tactile and force sensors, audio data, and the robot’s own physical state into a single unified understanding framework, allowing it to interpret both its surroundings and its own actions.

Neura argues that vision-only approaches represent an amateur approach. A genuinely useful robot must be able to “feel” whether a screw has engaged properly or “hear” the crisp click of a connector snapping into place—subtle details that cameras cannot capture. By combining multiple sensing modalities, the model aims to handle manipulation tasks that require precision and continuous feedback.

The “synthetic brain” is forged in simulated environments known as NEURA Gyms. There, robots learn by observing humans wearing sensor suits and repeatedly practice real-world tasks in simulation before deployment. This training setup allows the model to encounter varied scenarios without risking damage to physical hardware.

Notably, the model introduces a critical “memory layer” that gives robots object permanence—the understanding that an object continues to exist even when it is no longer in sight. This capability allows the AI to handle complex, multi-step tasks without losing track of what it is doing midway through an operation, a common failure mode for less sophisticated systems.