ALOHA 2 is an enhanced, low-cost hardware platform for bimanual teleoperation, designed to improve performance, ergonomics, and robustness compared to the original ALOHA system. The platform includes two ViperX 6-DoF arms (the "follower") and two smaller WidowX arms (the "leader"), with the leader arms being teleoperated by the user. The follower arms are controlled by backdriving the leader arms, allowing for efficient and intuitive teleoperation. The system features improved grippers with low-friction rail designs, enhanced gravity compensation, and a redesigned frame that allows for more space for human-robot interaction and props. The cameras have been upgraded to smaller, high-resolution RealSense D405 cameras, providing better depth and field of view. A MuJoCo model of ALOHA 2 has been released for simulation, enabling improved data collection and policy learning. The system also includes a user-friendly interface with ergonomic improvements, such as adjustable height chairs and swappable finger attachments. The platform supports large-scale data collection for fine-grained bimanual manipulation, enabling research in robot learning. The ALOHA 2 system is open-source, with detailed hardware designs and a tutorial provided for researchers. The system has been tested with users, showing improved performance and user experience compared to the original ALOHA system. The platform is designed to support a wide range of tasks, including industrial tasks with tight tolerances, and is expected to advance research in robot learning through its enhanced capabilities.ALOHA 2 is an enhanced, low-cost hardware platform for bimanual teleoperation, designed to improve performance, ergonomics, and robustness compared to the original ALOHA system. The platform includes two ViperX 6-DoF arms (the "follower") and two smaller WidowX arms (the "leader"), with the leader arms being teleoperated by the user. The follower arms are controlled by backdriving the leader arms, allowing for efficient and intuitive teleoperation. The system features improved grippers with low-friction rail designs, enhanced gravity compensation, and a redesigned frame that allows for more space for human-robot interaction and props. The cameras have been upgraded to smaller, high-resolution RealSense D405 cameras, providing better depth and field of view. A MuJoCo model of ALOHA 2 has been released for simulation, enabling improved data collection and policy learning. The system also includes a user-friendly interface with ergonomic improvements, such as adjustable height chairs and swappable finger attachments. The platform supports large-scale data collection for fine-grained bimanual manipulation, enabling research in robot learning. The ALOHA 2 system is open-source, with detailed hardware designs and a tutorial provided for researchers. The system has been tested with users, showing improved performance and user experience compared to the original ALOHA system. The platform is designed to support a wide range of tasks, including industrial tasks with tight tolerances, and is expected to advance research in robot learning through its enhanced capabilities.