NEURA Unveils Lighter 4NE1 Gen 4 as TUM RoboGym Opens in Munich

NEURA Robotics has unveiled 4NE1 Gen 4, a slimmer humanoid with a black textile exterior that the company says is 20% lighter than its predecessor. The October 22 reveal accompanied the opening of the TUM RoboGym at Munich Airport's LabCampus, linking the new hardware to a facility for teaching robots practical tasks.

Key Takeaways

Show all

• NEURA says 4NE1 Gen 4 is 20% lighter, with upgraded hands, limb-level reflexes and Neuron AI.
• The robot is being prepared for series production; the announcement does not specify pricing or customer delivery dates.
• TUM says its new RoboGym will initially train around 100 humanoid and service robots in 48 cells.
• A separate NXAI partnership aims to add stronger memory capabilities to NEURA's robotics software.

A new body and control system

In its launch announcement, NEURA describes improved gripping and “Smart Limbs,” with local reflexes allowing limbs to respond independently. The company says Gen 4 is already operating in its Gyms and is also appearing at NVIDIA GTC Berlin.

Neuron AI combines visual information with touch, force, sound and robot-state inputs to guide movement. NEURA presents the textile covering as a more approachable appearance with scope for customization. These are company descriptions; the release does not provide comparative manipulation benchmarks or an absolute weight for the new robot.

A training center for practical work

According to TUM's opening announcement, the RoboGym brings together 48 training cells and an initial population of roughly 100 humanoid and service robots. Those figures cover the facility's robot mix, rather than 100 units of the new 4NE1.

Tasks include folding boxes, assembling components and using tools. Data generated through physical interaction will feed a curated cloud data lake for research into robot skills and safe human-robot interaction. Industrial applications and internal logistics are the initial focus.

The opening follows our earlier coverage of the NEURA–TUM training-center project. It gives researchers and companies a shared place to test how well a learned skill survives contact with real objects and workflows.

European competitor Generative Bionics is pursuing a similar approach. As we reported on October 21, it plans a 5,000-square-metre Physical AI Data Factory in Italy, using AMD infrastructure to combine robot practice with simulation and model training. That facility is scheduled to become operational in Q4 2027, while the Munich RoboGym has now opened.

NXAI partnership adds a memory component

Two days before the launch, NEURA announced a partnership with Austria's NXAI to develop memory for cognitive robots. The work combines NEURA's Physical AI software with xLSTM, a recurrent architecture developed under Sepp Hochreiter that maintains and updates state across long sequences.

The aim is to help robots keep track of multi-step tasks and draw on previous experience. NXAI also plans to offer its own model through Neuraverse, alongside the independently developed Neuron AI. Companies would be able to adapt it, test it in NEURA Gyms and deploy it on suitable robots.

That partnership describes ongoing development, not confirmation that the new NXAI memory component is already running on Gen 4. Together, the announcements outline NEURA's approach: new robot hardware, facilities for collecting experience, and software intended to make that experience useful across longer tasks.