AGIBOT Worken: Measuring Robots by Useful Work, Not Units
Key Takeaways
- AGIBOT unveiled Worken at AGIBOT Innovation Day 2026 in Shanghai on October 29, 2026, as a proposed measure of robots’ useful output.
- Worken stands for “Effective Intelligent Work Hours”: active robots multiplied by autonomous working hours multiplied by an output efficiency ratio.
- Worken is a company proposal, not an adopted industry standard, and it still needs agreed benchmarks for each task.
- The same event introduced the AHA3 hardware platform, the GO-3 Preview embodied foundation model, the WITA-Omni interaction model, the G2 Max industrial robot and Link-U OS 1.0.
- AGIBOT’s argument is that robot value should be judged by work completed, not by how many units are built or shipped.
How should a robot be measured? By how many are built, by how many are sold, or by how much useful work they actually do? At AGIBOT Innovation Day 2026 in Shanghai on October 29, 2026, Chinese robotics company AGIBOT proposed an answer: a framework called Worken, designed to measure productive output once robots are deployed in the real world.
The idea is simple to state and hard to standardise. Worken combines three factors: the number of robots actively performing tasks, the hours they spend working autonomously, and an output efficiency ratio that scores their output against a reference benchmark for a defined task. A robot sitting in inventory counts for nothing. A robot powered on but idle counts for little. What counts is acceptable work completed, autonomously, against a stated standard.
What Worken actually measures
AGIBOT describes Worken as “Effective Intelligent Work Hours”. The name combines “Work” and “Token”, but the arithmetic is what matters. The first factor counts robots actively performing tasks in real-world environments, not units merely manufactured or delivered. The second measures hours spent working autonomously rather than total powered-on time. The third scores output against a reference benchmark for a defined task, such as the amount of acceptable work a human worker completes under established operating standards.
Peng Zhihui, AGIBOT’s co-founder, president and chief technology officer, presented the framework as an answer to a measurement question he said the industry must address as robots begin working in factories, warehouses and service environments. In the company’s announcement, he said the aim was to move beyond measuring how many robots are built or delivered and to begin measuring how effectively they can perform useful work, autonomously and consistently, in the real world.
AGIBOT also frames Worken as an economic tool. Hardware depreciates over time, but improvements in software and AI models may let the same machine perform more useful work during its operating life. Weighing output against total lifecycle costs could therefore help businesses judge whether a deployment is actually worth its price, rather than judging it by the purchase cost alone.
Why the industry might need a new yardstick
Robotics has a familiar reporting problem. Unit shipments and deployment announcements are easy to count, but they say little about reliability, autonomy or output quality. A fleet of robots that needs constant human help, or that completes tasks slowly and with frequent errors, may look impressive in a press release and disappointing on a factory floor. A measure based on autonomous working hours and benchmarked output is an attempt to close that gap.
The difficulty is agreement. Worken only becomes comparable across companies if everyone uses the same task benchmarks, the same definition of “autonomous”, and the same way of scoring acceptable output. AGIBOT said it views Worken as an evolving framework that will be refined through real-world deployments and operational data, and that it hopes the measure can eventually become a broadly applicable reference. That is an admission that the hard work — agreeing benchmarks task by task — is still ahead.
The same caution applies to the deployment examples AGIBOT cited. The company pointed to G2-family deployments in tablet production at Longcheer, semiconductor packaging and testing, CNC machine tending and logistics, with performance assessed against cycle time, sustained operation and successful task completion. Those are company-described deployments, not independently audited results, and they do not by themselves validate Worken as a universal measure.
The hardware and models behind the announcement
Worken was not the only announcement at the event. AGIBOT introduced AHA3, its third-generation humanoid hardware architecture, described as its first commercial “Three Intelligences in One” platform, combining the hardware foundations for locomotion, interaction and task execution intelligence. The platform builds on advances in motor-drive technology, integrated computing and Wi-Fi 6 and 5G connectivity, alongside a liquid-cooling system for robot joints, specialised joints with IP67-rated protection, a dexterous-hand family and the OmniPicker 3 tactile gripper.
On the model side, GO-3 Preview is a next-generation embodied foundation model built around WRAM, a unified framework for reasoning, prediction and action. AGIBOT said it demonstrated zero-shot and few-shot task capabilities, and, according to evaluation results the company presented at the event, it took first place in task success rate and process score on RoboDojo, the highest overall score on RoboTwin and first overall on RoboColiseum. Those are company-presented benchmark claims, not independent evaluations, and they should be read as such.
WITA-Omni, meanwhile, is an omni-modal interaction model bringing visual perception, speech, real-time dialogue, facial expressions and physical actions into a single continuous interaction system, supporting full-duplex conversation and on-device deployment. AGIBOT said it currently ranks first on multiple embodied-interaction benchmarks, including Daily-Omni — again, a company claim, not an independent finding.
G2 Max, Link-U OS and real-world use
The G2 Max is a heavy-duty industrial robot for manufacturing and logistics, with 21 degrees of freedom, a dual-arm handling capacity of up to 40 kilograms and an adjustable working height from 0.1 to 2.5 metres. AGIBOT said its force-controlled manipulation, omnidirectional mobility and safety features are intended to support efficient operations alongside existing industrial equipment. Beyond manufacturing, the company highlighted Guangzhou Metro and the Chimelong tourism and entertainment complex, where robots perform tasks such as guiding visitors and answering questions.
Link-U OS 1.0, which AGIBOT positions as an operating system purpose-built for embodied intelligence, was also formally released, built on a unified hardware abstraction layer, an embodied-intelligence-native kernel and an intelligent service framework. AGIBOT announced an OpenAtom Foundation Platinum Level Sponsorship and the donation of the Link-U OS project to the foundation, with plans for it to become fully open source by the end of 2026. It is currently being trialled by selected ecosystem partners, so that open-source promise is a plan, not yet a completed release.
The physical-AI context matters here. Robots are increasingly judged by what they do in warehouses and factories, not by laboratory demos, as shown by Otto and Sereact’s NVIDIA physical AI robots for returns sorting and by Bedrock Robotics’ autonomous excavators. Worken is AGIBOT’s attempt to give that judgement a number.
Limitations and unanswered questions
Worken is a proposal, not a standard. No independent body has adopted it, no cross-company benchmark set has been published, and no independent audit of AGIBOT’s deployment claims was available in the source reviewed. The GO-3 Preview and WITA-Omni benchmark claims are company-presented. The open-source release of Link-U OS is planned for the end of 2026, not yet delivered. Until task benchmarks are agreed and independently applied, Worken is best understood as a useful question — how much useful work did the robot actually do? — rather than a settled answer.
