AgiBot Releases Heavy-Load Embodied Robot Genie G2 Max, Already Handling Joint Transport on the Qiyuan T1 Production Line

After a few days away, the robots have "evolved" again. AgiBot recently released the safety-standard heavy-load embodied robot Genie G2 Max, which has already taken on the task of "building robots" on the personal robot Qiyuan T1 production line. At the same time, Qiyuan announced the rollout of its 5,000th robot, just one month after sales began. In addition, another robot in the AgiBot Genie family, the Genie G2, has been on duty for some time at a convenience store in Shanghai. Equipped with AgiBot's new embodied agent framework, it has the ability to "learn from its mistakes" like a human, and can independently locate the cause of failure, adjust strategies, and improve skills in an unmanned environment. This is the industry's first time stepping outside simulation environments and out of the laboratory to verify embodied recursive self-evolution capabilities in real business scenarios. The Genie G2 Max will soon be equipped with this "self-learning" skill package as well. Heavy load and precision combined, robots "build" robots. On the Qiyuan T1 production line, the Genie G2 Max undertakes a "robot builds robot" task. The Genie G2 Max has a single-arm payload of 20 kilograms and a dual-arm peak payload of up to 50 kilograms, with arm repeat positioning accuracy controlled within 0.5 millimeters, making it the industry's first heavy-load embodied robot to achieve long-range, multi-task operations. AgiBot releases a robot that "builds robots," with a steady rhythm and no drowsiness: assembling a "companion" in 2.5 minutes. As the "precision porter" on the joint assembly line, the G2 Max is responsible for accurately delivering limb joint components such as finished shoulder joint housings and hip joint assemblies from the material stacking area to the final assembly station. These components are the parts of the robot with the highest precision requirements. In the past, they were entirely reliant on manual box handling, cart pushing, loading and unloading; even a slight bump to the mounting surface could affect the service life of the joint module. After the G2 Max came on duty, the entire dual-arm end-box handling process delivers with sub-millimeter precision, completing pickup, movement, and delivery in one step. With its heavy-load and precision handling capabilities, it "lifts" the birth of another robot. The production takt of the Qiyuan T1 is one robot rolling off the line every 2.5 minutes. Zhou Ye, factory director of the Huguang Co. robot manufacturing plant, said this means the material supply rhythm for the entire joint assembly station must be perfectly coordinated; if it is late, downstream stations have to wait idle. "The G2 Max's operating takt is very stable, and its all-day takt curve is almost a straight line," he said. Equipped with dual-battery hot-swapping and autonomous recharging multiple energy replenishment solutions, the G2 Max is also unafraid of night shifts and can achieve 7×24-hour uninterrupted operation. Designed according to industrial safety standards, it builds 360-degree multimodal perception protection, integrating 2D/3D LiDAR, ultrasonic sensors, chassis safety bumpers, and physical emergency stops, so it can operate safely even in complex industrial environments. Flexible operations across the full area, no production line modification needed. In tasks such as material box transfer, production line feeding, and heavy component pickup and placement, robots need to simultaneously meet three requirements: carrying heavy objects steadily, placing materials accurately, and keeping up with multi-shift production. Especially under heavy-load conditions, robots still need to maintain precise pickup and placement capabilities and continuously complete repetitive tasks. AgiBot stated that the G2 Max does not simply pursue the ceiling of load parameters, but instead invests substantial engineering capability in force control and vision fusion solutions to solve the long-standing problem of "able to carry but unable to place accurately." Relying on 7-degree-of-freedom force-controlled dual arms, combined with visual guidance and dual-arm force control technology, the G2 Max is compatible with various industrial materials such as cardboard boxes, European-standard turnover boxes, and irregular components, without requiring extensive re-teaching and calibration. AgiBot releases a robot that "builds robots," with a steady rhythm and no drowsiness: assembling a "companion" in 2.5 minutes. The biggest obstacle for many factories in advancing automation is the high cost of production line and site modification. Once heavy-load equipment enters the site, workstations often have to be torn down and rebuilt, and large movement spaces reserved, keeping the deployment threshold high. The G2 Max takes "compatibility with existing production lines" as a core design goal. Taking material pickup as an example, in the past a robotic arm needed two motions to complete the task. The G2 Max is equipped with a high-precision waist pitch joint and can actively lean forward to adjust its body posture, achieving single-side depalletizing and palletizing. Pallets can be placed directly against the wall, with no need to reserve 1.5 meters of space behind for maneuvering, greatly freeing up factory floor area. This spatial adaptability has already been applied in leading e-commerce warehousing scenarios. The G2 Max can seamlessly connect to the stereoscopic warehouse "goods-to-person" system and collaborate with existing robot systems to complete the full process of full-box depalletizing and inbound storage, as well as empty material box palletizing and recovery. Its overall operation success rate exceeds 99%, and it can depalletize 80 boxes and palletize 75 boxes per hour, completing automation upgrades directly within the original warehouse layout. Learning on the job, continuous evolution in unmanned environments. After robots truly "take up their posts," they will inevitably encounter many scenarios not seen during training. In addition to the Genie G2 Max, AgiBot also released the embodied agent framework "Genie Master" at the same time, enabling robots to unlock new skills while on duty, just like humans. Over the past period, the Genie G2 has quietly been on duty at a convenience store in Shanghai and has achieved continuous improvement in multiple complex and delicate tasks: cooked food packaging has become smoother after repeated failures, and the accuracy of pickup and restocking actions has also steadily improved. These changes do not come from manual parameter tuning, but from the robot independently locating problems and accumulating experience during execution. This benefits from two recursive routes: one lets the robot remember experience, and the other lets the robot turn experience into capability. Taking takeout packaging as an example, to address the problem that water bottles and grippers are easily entangled by white carrying straps, the agent uses more specific action prompts to guide the gripper to retract first and then try again to put the item into the packaging bag. Such execution and recovery data then flows back into the training process, becoming material for subsequent action learning. The robot thus learns to "learn from its mistakes." AgiBot also offered an early "spoiler": in a few days it will launch a 24-hour uninterrupted livestream. Six Genie G2 robots will serve as "probationary employees" at the robot convenience store, going through four shifts: the lunchtime peak trial shift, the evening peak stress test, late-night independent duty, and the early-morning breakfast rush. Will they successfully "become regular employees"?