Munich Startup RLWRLD Revolutionizes Robotics with Open-Source AI and Marketplace
Munich, June 23, 2026 – RLWRLD, a Munich-based startup, has set new standards in the field of ‘Physical AI’ with the release of its open-source AI model, RLDX-1, and the launch of the ‘All Hands Up!’ marketplace for robotic hands. This development underscores Germany’s growing influence in the robotics sector and has attracted billions in investments.
RLDX-1: An Open-Source Competitor for Dexterous Robotic Hands
In May 2026, RLWRLD unveiled RLDX-1, a Foundation Model boasting 8.1 billion parameters. Its open-source nature allows any developer to utilize and further develop it, fostering a collaborative environment for innovation. The model was trained on Amazon Web Services (AWS) cloud infrastructure, specifically on EC2 p5e and p5en instances equipped with NVIDIA H200 GPUs.
The performance of RLDX-1 has been remarkable. It achieved top scores on leading benchmarks such as RoboCasa, LIBERO, and SIMPLER. In the SIMPLER-Google-VM test, it reached an impressive success rate of 81.5 percent, utilizing approximately 20 percent less computing power than previous development cycles. The model outperformed competitors like GR00T N1.6 on eight different benchmarks.
“All Hands Up!”: An Open Marketplace for Robotic Hands
On June 22, 2026, RLWRLD launched “All Hands Up!”, a platform offering technical reports, interactive simulations, and data on over ten different robotic hands. The platform distinguishes between “Type 1” hands for industrial manufacturing and “Type 2” hands designed for precise data acquisition and machine learning.
A core component of the platform is DexBench, a test catalog developed in collaboration with NVIDIA. It comprises 18 tasks across five evaluation categories and establishes a 5-finger data standard for humanoid robots. DexBench will be integrated into the NVIDIA Isaac platform, creating a shared evaluation infrastructure for developers worldwide.
From Seoul Hotel to Precision Robotics
To enhance the data foundation for fine motor skills, RLWRLD forged an unconventional partnership with the LOTTE HOTEL & RESORT in Seoul. A joint analysis identified over 400 tasks and 4,500 specific hand movements requiring 5-finger dexterity, transforming the hotel industry into a training ground for the next generation of robots.
Safety First: NVIDIA’s “Halos” for Autonomous Machines
The “Physical AI” industry is also prioritizing safety. On June 22, 2026, NVIDIA introduced “Halos for Robotics,” a safety system for autonomous machines. Built on IGX-Thor hardware and the Halos operating system, it complies with the IEC 61508 SIL 3 standard, the highest safety level for industrial controls. Partners include Agility Robotics and Boston Dynamics.
Concurrently, the Automate 2026 trade fair in Chicago showcased the industry’s focus on safety. Mantis Robotics presented the MR-X, a biomimetic dual-arm robot with a 31.7-kilogram payload capacity and speeds of up to 10.6 meters per second. Cobot launched the Proxie Gen2, which features 40 percent fewer components and can move carts weighing up to 680 kilograms.
Billions for Physical AI
The development of Physical AI is attracting significant investment. In 2026 alone, $1.03 billion flowed into a seed round for AMI Labs, and approximately $1 billion into World Labs. RLWRLD itself secured a seed extension of $41 million, while Limitless Labs is advancing automated CNC programming with a $20 million Series A investment.
Heavy industry is also joining the trend. ArcelorMittal announced a strategic partnership with AWS on June 22, 2026. The steel giant plans to implement computer vision and digital twins in its plants across 14 countries. In return, Amazon will procure lower-emission XCarb steel for its data centers, illustrating the close intertwining of AI and industry.
Source: boerse-global.de
