German robot maker NEURA Robotics said on 14 October 2026 that it has launched Neuron AI, which it calls “its multimodal foundation model for Physical AI.” In plain terms, it is a single AI model meant to act as the “brain” for NEURA’s robots, turning what they sense into physical actions.
Everything below about what the model can do comes from NEURA’s own announcement. No outside group has yet tested it.
What NEURA says Neuron AI does
A foundation model is a large AI model trained on broad data that can then be adapted to many tasks. Most robot AI relies mainly on cameras. NEURA says Neuron AI goes further by “combining 3D vision, touch, audio, robot state and language with memory.” Robot state means the robot’s own information, such as the position of its joints.
NEURA’s argument is that cameras miss a lot. Its release gives examples: when a robot opens a lock, only sound and touch confirm that it is really open, and with touch a robot can grasp objects it cannot see. The company also says the sound of welding can show the quality of a seam.
NEURA calls its design a “Perception-Language-Action (PLA)” architecture: it links what the robot senses with instructions and with what it has already done, and turns that into movement.
The model also has memory. NEURA says a robot needs “object permanence”, the ability to remember that things still exist when it cannot see them, and to keep track of what its own actions have done over time.
How it is trained
NEURA says Neuron AI learns from three kinds of data:
- real robots doing tasks in NEURA’s training facilities, which it calls NEURA Gyms, recording vision, touch, force, sound and movement;
- simulation, using a tool NEURA calls its Data Amplifier to create many variations of tasks and settings; and
- human demonstrations, recorded by people wearing sensor-equipped suits.
NEURA says the same model works across five of its robot types, with different grippers and hands. It says the model was “developed and trained in-house in Metzingen, Germany,” where the company is based, and that it has data partnerships with industrial companies “including Schaeffler and Bosch.”
How customers would use it
According to NEURA, customers can fine-tune the model for their own jobs and send the resulting skills to fleets of robots through its Neuraverse platform. The company says customers “can keep their skills private,” or choose to share them to help improve future versions. NEURA says its key measure of progress is “Time-to-Skill”: how quickly a robot can learn a new task and do it reliably.
A laundry demonstration
NEURA plans to show its humanoid robot, 4NE1, carrying out what it calls “an eleven-minute laundry-handling sequence” at NVIDIA’s GTC Berlin conference from 20 to 22 October. The company says the robot will sort laundry, load a washing machine, close and reopen it, unload it and move the laundry to a dryer “without human intervention.” The demonstration has not taken place yet, and the claim that it will run fully on its own is NEURA’s.
“With Neuron AI, we are building the brain of the physical world,” said David Reger, NEURA’s founder and chief executive. “Developed in Germany, built for the world.”
What this means
- Robot makers are building their own AI brains. NEURA is betting that touch, sound and memory, not just vision, are needed for robots to do real-world work.
- The claims are untested. NEURA has published no independent benchmarks, and it has not said when the model will be widely available or what it will cost.
- The demo is the first public test. The planned laundry task at GTC Berlin will be the first chance to see the model work in public.
Sources
- NEURA Robotics via Business Wire, “NEURA Robotics Introduces Neuron AI: The Brain of the Physical World”, 14 October 2026. https://www.businesswire.com/news/home/20261013482085/en/
Related stories
- Industrial Robots This Week: An Ex-Optimus Startup, AWS’s Robot Toolchain and Who Leads the Warehouse
- NVIDIA’s “Physical AI” Play: Teaching Robots to Think, Talk, and Actually DO Things
- Shenzhen Robot Maker Astribot Raises About RMB 1bn at a RMB 15bn Valuation, Its Fourth Round in About Six Months
- Boston Dynamics Names Rohit Prasad CEO as Atlas Heads for the Factory
