NEURA Robotics and NXAI announced a partnership on October 20, 2026 to jointly develop a memory component for cognitive robots, integrating NEURA’s Physical AI stack with NXAI’s xLSTM architecture and adding a second model option to the Neuraverse platform.
A Two-Part Partnership
The partnership, datelined Metzingen, Germany, consists of two parts, according to the announcement. The joint work builds on Neuron AI, NEURA’s multimodal foundation model for Physical AI, which combines 3D vision, touch, audio, robot state and language with a memory function. The partners are integrating NEURA’s Physical AI stack with xLSTM, which the announcement describes as a recurrent architecture from NXAI that can hold and update state efficiently across long sequences. NXAI also plans to make its own xLSTM-based model available on the Neuraverse platform, where it would sit alongside Neuron AI; the announcement presents this as a plan rather than a completed availability.
“A robot that genuinely helps in everyday life and industry must be able to remember,” David Reger, founder and CEO of NEURA Robotics, said in the announcement. Reger described Neuron AI as the foundation, the Neuraverse as the platform through which outside AI approaches can find their way into robots, and NXAI as the partner for building out that memory.
Prof. Dr. Sepp Hochreiter, founder of NXAI, said memory and experience are what allow robots to solve tasks reliably across many steps, and that NXAI and NEURA together are bringing that strength into Physical AI.
Working and Episodic Memory
The announcement contrasts robots with chatbots to explain why memory matters for physical machines. A chatbot processes requests one at a time, while a robot operates in a continuous loop of perceiving, deciding and acting many times per second, potentially for an entire shift. The release distinguishes two levels of memory: working memory handles the immediate state of a task, such as whether a component has already been inserted, whether a door is open or closed, or whether a start button has been pressed. Episodic memory goes further, allowing a robot that has performed a process repeatedly to draw on prior experience when it encounters the same situation again.
Neuraverse Integration and Data Control
Under the planned integration, companies would be able to fine-tune the NXAI model for their own tasks using the Neuraverse toolchain, test it in the NEURA Gyms and deploy it on suitable robots, all within the same training and application environment they already use, according to the announcement. Neuron AI remains NEURA’s own multimodal foundation model and will continue to be developed independently; the partnership adds a second model option to the platform and includes joint research. Companies retain control over their process knowledge and decide for themselves whether to share data and skills.
What Each Partner Contributes
NEURA contributes its Physical AI stack: proprietary multimodal foundation models, robotics, sensor technology, real-world training data from the NEURA Gyms and the platform for industrial deployment. NXAI brings what the announcement describes as decades of AI research and a fundamentally different approach to recurrent intelligence and memory.
NXAI is an Austrian deep-tech company based in Linz, and its core research focuses on xLSTM models for time series and robotics. On its own site, the company describes itself as a frontier lab and says its proprietary xLSTM memory architectures enable state-tracking and outperform the currently dominant Transformer architectures. Hochreiter is NXAI’s co-founder and Chief Scientist, and the company says it works closely with JKU Linz and other research institutions. NXAI also says it is building specialized world models based on time-series analysis under the name TiRex, and that it will develop large-scale world models built on neural memory hierarchies and memory routing, with applications in robotics.
NEURA Robotics was founded in 2019 and is headquartered in Metzingen. The company develops cognitive robots that can see, hear, feel and learn, along with the software, AI and data infrastructure required to deploy them at scale, and it describes the Neuraverse as an open ecosystem in which robots continuously learn and share skills.
Neuron AI, according to NEURA’s product documentation, is a Perception-Language-Action model that encodes 3D vision, tactile and force-torque data, audio, robot state and language in one shared space over time and turns them into a single action stream. The model runs locally on the robot without a cloud connection and was developed and trained in-house in Metzingen. Training draws on three sources: real-world robot data, simulation-amplified data and human demonstration data, complemented by data partnerships with industrial companies including Schaeffler and Bosch. NEURA says Neuron AI runs on five robot embodiments: the 4NE-1 Gen 3 and Gen 4, a wheeled 4NE-1, and two further bimanual platforms.
The partners’ stated longer-term ambition is for robots to build on experience over time rather than only execute individual skills. The announcement ties that goal to the vision behind Neuraverse and what NEURA calls the Robot Skill Economy, in which skills acquired through real-world interaction are refined and carried across machines and environments.
Credit: Source link
