NEURA Robotics and NXAI partner on xLSTM memory for physical AI

NEURA Robotics and Austrian AI company NXAI have partnered to develop memory for cognitive robots, combining NEURA’s Physical AI technology stack with NXAI’s xLSTM recurrent architecture. The 20 October agreement covers a jointly developed memory component and plans for a separate NXAI model on Neuraverse, NEURA’s robot-training and deployment platform.

The work targets robots performing extended sequences of physical actions. Working memory would track immediate process state, such as whether a component has been inserted or a door opened, while episodic memory would allow the system to draw on previous task experience. NXAI’s architecture, developed under founder Sepp Hochreiter, is designed to retain and update state efficiently across long sequences.

Companies using the planned NXAI model would be able to fine-tune it for their own tasks through Neuraverse, test it in NEURA Gyms and deploy it on suitable robots. NEURA will continue developing Neuron AI independently, with the partnership adding another model option and joint research to the platform. Customers will decide whether to share their process data and learned skills.

The agreement follows NEURA’s 14 October introduction of Neuron AI, a Perception-Language-Action model combining 3D vision, touch, audio, robot state and language with memory. NEURA says the model runs across five robot embodiments and uses real robot interactions, simulation and human demonstrations for training. Its April collaboration with Amazon Web Services established AWS as Neuraverse’s primary cloud provider and connected SageMaker with the training pipeline.