AgiBot has proposed a way to measure how much useful work deployed robots actually do, while unveiling new robot hardware, AI models and software. The Worken metric moves the conversation beyond counting machines, though the company says it is still developing the measure using real-world deployment data.
Watch Desk analysis
What happened
At its Innovation Day in Shanghai, AgiBot defined Worken as the number of active robots multiplied by autonomous working hours and an output-efficiency ratio. That ratio compares completed work with a reference benchmark for a defined task. The company says the framework is intended to help compare productivity across uses such as manufacturing, logistics, inspection and commercial services. The AI Insider reports that AgiBot plans to refine it with data from deployments.
The company also introduced AHA3, a third-generation humanoid hardware architecture; GO-3 Preview, an embodied-AI model intended to combine reasoning, prediction and action; and WITA-Omni, an interaction model for vision, speech and movement. AgiBot formally launched its G2 Max industrial robot, with 21 degrees of freedom and a stated dual-arm handling capacity of up to 40 kilograms.
AgiBot released Link-U OS 1.0 to selected partners for testing and said it plans to make the operating system fully open source by the end of 2026. The company also said it had donated the project to the OpenAtom Foundation.
Why it matters
Robot makers have plenty of ways to count machines and announce capabilities. A measure tied to operating hours and task output could make it easier for customers to ask a more useful question: what work is a robot actually completing? But the benchmark behind the efficiency ratio will matter, and AgiBot’s proposal is not yet an established industry standard.
The announcements also show the breadth of AgiBot’s pitch, from hardware and embodied-AI models to the software layer connecting robots with applications. Those are company-described capabilities and plans; real deployments and comparable results will show how much they deliver.
Our read
Worken is the most interesting announcement because it puts productivity, rather than robot-count theatre, at the centre of the discussion. The formula is a starting point, not proof that different robots can already be compared fairly. For operators, the useful questions are what counts as output, who sets the reference task and whether the numbers hold up outside a company’s own demonstrations.
What to watch
- How AgiBot defines reference benchmarks and validates Worken with deployment data.
- Whether other robotics companies adopt or challenge the proposed measure.
- What independent results show for GO-3 Preview and the G2 Max.
- Whether Link-U OS becomes fully open source by the end of 2026.
Discussion spark: Should robot productivity be measured against human task benchmarks, or should the industry build a different standard for work done by machines?
Sources and evidence
- AgiBot Introduces Worken Metric to Measure Productive Robot Labor (29 October 2026, 13:48 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.