NEURA Robotics launches Neuron AI, a robot model that hears and feels

NEURA Robotics launches Neuron AI, a robot model that hears and feels

14 October 2026: NEURA Robotics, based in Metzingen, Germany, has introduced Neuron AI, a multimodal foundation model for its robots that combines 3D vision, touch and force, audio, robot state and language with a memory of what has happened. The company says it lets robots act on information cameras cannot capture and keep track of objects that have moved out of view.

Not yet independently verified. NEURA has published no benchmark or Time-to-Skill figures and no price or general availability date; the capability claims are its own. We will update this when it can be confirmed, and remove this note.

How the model works

NEURA calls the architecture Perception-Language-Action. Rather than treating each sensor separately, the model encodes all inputs into one shared space over time and turns them into a single stream of actions. The company's examples are tasks vision cannot settle on its own: only sound and touch confirm that a lock has opened, touch lets a robot grasp something it cannot see, and in welding the sound of the process indicates the quality of the seam. The memory function gives the robot what NEURA calls object permanence, the ability to remember things it can no longer see.

"An intelligent robot needs a brain that brings its senses together and remembers what happened," said David Reger, founder and chief executive.

Where the training data comes from

NEURA trains the model on three sources:

• real robot data from its NEURA Gyms, its training facilities for robots, with synchronised vision, touch, force, audio and movement;
• simulated variations of tasks and environments from its Data Amplifier and simulation stack;
• human demonstrations captured by people wearing sensor-equipped data suits.

Data partnerships with Schaeffler and Bosch add to the gym data. The model runs across five NEURA robot embodiments with different grippers and hands, and was developed and trained in-house.

What customers get

Customers fine-tune Neuron AI for their own tasks and roll skills out across robot fleets through Neuraverse, NEURA's platform, keeping skills private or sharing them. The company's measure of progress is Time-to-Skill, how quickly a robot learns a new task and performs it reliably, though it has not published figures. At NVIDIA GTC Berlin from 20 to 22 October, its 4NE1 humanoid is due to run an eleven-minute laundry sequence, from sorting to loading the dryer, with no human help. NEURA was founded in 2019. For manufacturers, the question to ask is how much demonstration data a new task takes and who owns the resulting skill. Our robot manufacturer category lists 57 companies.

Company profile on TrustList: NEURA Robotics

Sources

• NEURA Robotics: NEURA Robotics introduces Neuron AI, 14 October 2026