What Is NEURA's Neuron AI?

Neuron AI is NEURA Robotics' multimodal foundation model for physical AI, built on what NEURA calls a Perception-Language-Action architecture. NEURA introduced it on 14 October 2026 in a [press release titled "NEURA Robotics Introduces Neuron AI: The Brain of the Physical World"](https://neura-robotics.com/neura-robotics-introduces-neuron-ai/). The [Neuron AI product page](https://neura-robotics.com/neuron-ai/) gives the same name.

NEURA says Neuron AI combines 3D vision, touch, audio, robot state and language with memory, "so robots can react to what they cannot see". NEURA says five input streams go into one shared space and that Neuron AI turns them into a single stream of action. NEURA says customers fine-tune Neuron AI and deploy skills across robot fleets through its Neuraverse platform, and NEURA publishes no standalone price, licence or download for Neuron AI at the time of writing.

NEURA's Neuron AI page breaks the five inputs down by data type.

  • 3D vision: point clouds and depth, which describe the scene as space
  • Touch and force: force-torque and contact data from hands and grippers
  • Audio: contact sounds and spoken instructions
  • Robot state: joint positions, velocities and end-effector pose
  • Language: task prompts and physical concepts

How Neuron AI Differs from a VLA Model

NEURA says a vision-language-action (VLA) model sees the world through cameras and judges by what the camera sees, reasoning mostly on the current frame. Neuron AI "adds touch, sound and memory" to that camera view, according to the Neuron AI page. Readers new to the category can start with the guide to [embodied AI](/what-is-embodied-ai/).

NEURA defines the term directly: "A PLA model connects what a robot sees, hears and feels with what it has already done." NEURA gives three examples: a camera cannot tell whether a lock has opened, touch guides the grip when an object is hidden, and in welding the process sound reveals seam quality.

According to the press release, Neuron AI keeps track of objects that are out of sight (object permanence). NEURA says it adapted the architecture specifically to add memory and trains Neuron AI on dedicated datasets that use it.

Where Neuron AI Runs and What It Learns From

NEURA states that "Neuron AI is deployed locally", running real-time inference on the robot itself with no cloud round trip. According to the Neuron AI page, Neuron AI keeps working without a cloud connection. NEURA also says production data stays under the customer's control.

NEURA says it trained Neuron AI across five of its own robot embodiments, each with different grippers and hands. The five are the 4NE-1 Gen 3, the 4NE-1 Gen 4, a wheeled 4NE-1 and two further bimanual platforms.

NEURA describes three data pillars behind Neuron AI. NEURA also says data partnerships with industrial companies including Schaeffler and Bosch complement the training it runs in the NEURA Gyms. NEURA says it developed and trained Neuron AI in-house in Metzingen, Germany. Founder and CEO David Reger is quoted in the release: "An intelligent robot needs a brain that brings its senses together and remembers what happened."

  • Real-world robot data from NEURA Gyms and real environments, with synchronized vision, touch, force, audio and movement
  • Simulation-amplified data from NEURA's Data Amplifier and simulation stack
  • Human demonstration data from people wearing sensor-equipped data suits

Fleet Deployment Through Neuraverse

Neuraverse is the platform through which NEURA says customers fine-tune Neuron AI for their own tasks and deploy the resulting skills across robot fleets. Customers can keep those skills private. Where a customer chooses to share skills or deployment data, NEURA says that sharing can support the wider ecosystem and future model development.

NEURA names "Time-to-Skill" as its central measure of progress for Neuron AI. NEURA defines it as how quickly a robot learns a new task and then performs it reliably. At the time of writing, NEURA has published no benchmark scores, model size, parameter count or latency figures, so a reader cannot yet attach a number to that measure.

Readers weighing NEURA can take their questions to the [Neuron AI page](https://neura-robotics.com/neuron-ai/), which closes with "Talk to our team about Neuron AI on your robots, your use case and your data." Useful questions include which tasks run reliably at customer sites today, what fine-tuning through Neuraverse costs, and how the choice to share skills or deployment data is set up for a new customer.

The 4NE1 Laundry Demo at GTC Berlin

NEURA says it will demonstrate its humanoid 4NE1 ("For Anyone") performing an eleven-minute laundry sequence without human intervention at NVIDIA GTC Berlin 2026, using the current version of Neuron AI. The demonstration is listed for Station-Berlin, Hall 3, Booth #3031, from 20 to 22 October 2026. Background on the robot category sits in the guide to [humanoid robots](/what-are-humanoid-robots/).

According to the press release, the sequence covers sorting laundry, loading a washing machine, closing and reopening it, unloading it and transferring the laundry to a dryer. NEURA says the demo "focuses on handling and task coordination within the laundry process".

The GTC Berlin run is a NEURA-run demonstration, and it remains a claim until people outside NEURA have seen it. NEURA has not published which tasks work reliably at customer sites.

Details NEURA Has Not Published

At the time of writing, NEURA's Neuron AI page and press release leave the questions in the list below unanswered.

Funding context lives on the [NEURA Robotics company profile](/companies/neura-robotics/): [NEURA's 10 June 2026 announcement](https://neura-robotics.com/record-series-c/) puts its Series C at up to 1.4 billion dollars, naming Tether, Qualcomm Technologies, Amazon and NVIDIA among others, with no published valuation.

A buyer cannot price or schedule Neuron AI yet, since NEURA publishes no cost, licence or delivery date. The 20 to 22 October run in Berlin is the next dated event, and independent accounts of it would add evidence that NEURA's own release cannot supply.

  • A price for Neuron AI or for fine-tuning through Neuraverse
  • A licence or download option
  • A general-availability or delivery date
  • Benchmark scores, including any Time-to-Skill figure
  • Model size or parameter count
  • Latency figures for on-robot inference
  • Which tasks work reliably at customer sites today, or any independent test

Bottom Line

Neuron AI is NEURA's on-robot PLA model, and at the time of writing NEURA has published a product page, a press release and a scheduled demo for it, with no price, delivery date or benchmark. Readers who need a price, a delivery date or benchmark numbers to make a decision should not commit budget or plans around Neuron AI yet, because NEURA has published none of them. Readers following NEURA's humanoid work can see the laundry run at Station-Berlin, Hall 3, Booth #3031 from 20 to 22 October 2026 if they attend GTC Berlin. The verdict would change if NEURA published pricing and availability through Neuraverse, released benchmark or Time-to-Skill figures, or if an independent party tested 4NE1 outside NEURA's booth.

The [NEURA Robotics company profile](/companies/neura-robotics/) covers NEURA's Series C round and its robot lineup.

FAQs

What is a Perception-Language-Action model?

A Perception-Language-Action (PLA) model, as NEURA defines it, "connects what a robot sees, hears and feels with what it has already done." NEURA says Neuron AI encodes 3D vision, touch and force, audio, robot state and language in one shared space over time. It then turns those inputs into a single stream of action for the robot.

How is Neuron AI different from a VLA model?

NEURA says Neuron AI adds touch, sound and memory to the camera view that vision-language-action (VLA) models rely on. According to NEURA, VLA models judge by what the camera sees and reason mostly on the current frame. NEURA's example is a lock: a camera cannot tell whether it has opened, but sound and touch can.

Does Neuron AI need the cloud?

No, NEURA says Neuron AI does not need the cloud to run. Its Neuron AI page states that "Neuron AI is deployed locally" for real-time inference on the robot, with no cloud round trip. NEURA also says Neuron AI keeps working without a cloud connection and that production data stays under the customer's control.

Which robots does Neuron AI run on?

NEURA says it trained Neuron AI across five of its own robot embodiments with different grippers and hands. They are the 4NE-1 Gen 3, the 4NE-1 Gen 4, a wheeled 4NE-1 and two further bimanual platforms. NEURA has published no list of supported hardware beyond those five embodiments. Questions about other hardware go to NEURA through the [Neuron AI page](https://neura-robotics.com/neuron-ai/), which ends with "Talk to our team about Neuron AI on your robots, your use case and your data."

Where was Neuron AI developed?

NEURA says it developed and trained Neuron AI in-house in Metzingen, Germany, according to its 14 October 2026 press release. NEURA was founded in 2019, and David Reger is its Founder and CEO. Training also draws on data partnerships with industrial companies including Schaeffler and Bosch, which NEURA says complement its work in the NEURA Gyms.

Primary Sources