AgiBot Releases GO-3 Preview Embodied Foundation Model, First to Propose the Worken Workload Metric
AgiBot has released its new-generation embodied foundation model GO-3 Preview, which adopts its self-developed WRAM architecture to integrate task understanding, future state prediction, and action generation into a unified model, and is the first to propose the Worken evaluation framework for measuring a robot's actual work output. AgiBot says the model ranks first overall in three embodied intelligence benchmarks: RoboDojo, RoboTwin, and RoboColiseum. Worken equals the product of the number of active robots, autonomous working hours, and the output efficiency ratio; AgiBot also released the AHA3 hardware platform and Lingqu OS 1.0 at the same time.
Why read this
AI score 64
AgiBot releases the embodied foundation model GO-3 Preview and is the first to propose the Worken metric, which measures robots by actual work output; details on its benchmark performance and commercialization data are available.
Topics
Body · AI translation
English text is being prepared.
On October 29, AgiBot officially released its new-generation embodied foundation model GO-3 Preview, and for the first time proposed the Worken evaluation framework for measuring the actual work output of robots. During the same period, AgiBot also released the AHA3 "three-intelligence-in-one" hardware platform and Lingqu OS 1.0 and other technological achievements.
According to the introduction, GO-3 Preview adopts AgiBot's self-developed WRAM architecture, integrating task understanding, future state prediction, and action generation into a unified model, enabling robots to predict action outcomes based on environmental changes and generate corresponding operational actions, improving cross-task generalization and fine manipulation capabilities.
In terms of the training system, GO-3 Preview combines the Genie Evolver post-training system with the Genie Master embodied agent framework, integrating real robot, simulation environment, and world model data to build a continuous learning process covering pre-training, post-training, and actual deployment. AgiBot stated that this system aims to help robots transform experience from real operations into reusable task capabilities.
In terms of performance, AgiBot disclosed that GO-3 Preview has achieved first place overall in three embodied intelligence benchmarks: RoboDojo, RoboTwin, and RoboColiseum.
Notably, AgiBot this time for the first time proposed the concept of Worken, attempting to shift the evaluation focus of the robotics industry from device shipment volume to actual work output. According to the calculation method announced by AgiBot, Worken equals the product of the number of active robots, autonomous working hours, and output efficiency ratio, reflecting the ability of robots to continuously complete effective tasks.
On the commercialization side, Peng Zhihui, co-founder, president, and CTO of AgiBot, revealed that in the first half of 2026 the company shipped 9,700 robots, with overall revenue approaching 2 billion yuan, and in September the 20,000th robot rolled off the production line.
Submissions/leaks: tougao@youxituoluo.com
Article/business cooperation: Lin Nan (WeChat 19250561593)
Join the industry exchange group: Lin Nan (WeChat 19250561593)
Digital technology industry information and service platform
English text is being prepared.