AnswersPublished26 sources
X Square Robot Beat Figure on Speed—but Not on the Benchmark That Matters
X Square Robot reported 1,816 parcels per hour with more than 98% success in an unedited one hour test—about 45% above Figure’s 1,247.8 package per hour average—but the demonstrations were not comparable head to head... X Square used a stationary dual arm logistics cell; Figure used three bipedal humanoids over 200...
Published byEdited with GPT-5.6 LunaImages generated with GPT Image 1.5
Research answer
Key takeaways
- X Square Robot reported 1,816 parcels per hour with more than 98% success in an unedited one hour test—about 45% above Figure’s 1,247.8 package per hour average—but the demonstrations were not comparable head to head...
- X Square used a stationary dual arm logistics cell; Figure used three bipedal humanoids over 200 hours.
- The unresolved commercial metric is cost per correctly handled parcel in an independently observed customer operation, including interventions, retries, damage, uptime and maintenance.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: How did Shenzhen-based embodied-AI startup X Square Robot’s August 12, 2026 unedited one-hour livestream—during which its stationary dual-ar. Article summary: X Square’s stream was a strong demonstration of fast parcel induction, not a clean victory over Figure. X Square showed higher aggregate hourly throughput in a one-hour test—1,816 versus Figure’s 1,247.8-package-per-hour. Topic tags: general, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fa
X Square Robot’s August 12, 2026 livestream delivered an eye-catching number: 1,816 parcels processed in one unedited hour, with a stated success rate above 98%. That is roughly 45% higher than the 1,247.8 parcels per hour implied by Figure AI’s 249,560-package, 200-hour autonomous demonstration. 1
2
But the result is not a clean victory over Figure. X Square showed how fast a fixed, purpose-built dual-arm cell can run a parcel-induction task. Figure showed that humanoid robots could sustain autonomous logistics work for far longer. Those are important achievements—and different tests.
The headline numbers
X Square’s system used two stationary six-axis arms and simple grippers to identify parcels, pick them up, reorient them and feed them into a conveyor and scanning workflow. The company said the system handled parcels with varied sizes, shapes, materials and weights, including soft packages and boxes requiring orientation adjustments. 3
4
Its reported total of 1,816 parcels in 60 minutes works out to about 1.98 seconds per parcel. The visible target was 1,248 parcels per hour, almost exactly Figure’s reported average from its longer demonstration.
Figure’s Figure 03 robots performed a related logistics task: picking packages, identifying or orienting their barcodes and placing them on a conveyor. Early coverage of the livestream described a pace of roughly 2.6 seconds per package, while the final 249,560 packages over 200 hours equate to an aggregate average of 1,247.8 packages per hour, or about 2.89 seconds per package. 5
6
The arithmetic is therefore real. The comparison is not equivalent.
Why the machines were solving different problems
A fixed industrial cell can be designed around one conveyor, one work envelope and one high-volume process. Its arms do not need to walk, balance, navigate around obstacles or carry their own power systems. Short, repeatable trajectories can be optimized for speed.
A humanoid robot accepts additional mechanical complexity in exchange for mobility and compatibility with environments built for people. Legs, balance, onboard power, safety systems and a broader reachable workspace can make the platform more adaptable, but they also create costs and performance constraints that a fixed cell avoids.
That distinction is central to the business case. A parcel facility with a predictable layout and high volume may benefit more from a specialized dual-arm station. A customer that cannot redesign its site—or wants one machine to move among different work areas—may value the humanoid form more highly.
What the WALL-B demonstration shows
X Square presented the logistics system as a combination of its WALL-B embodied-AI model and self-developed six-axis robotic arms. The public demonstration showed the system handling varied parcels and recovering visibly after a routing or handling error before continuing. 3
7
That recovery is useful evidence of a closed-loop system: perception, action and correction appeared to operate together rather than stopping at the first unexpected event. It is stronger than a perfectly clean, edited clip.
It still does not prove that the underlying capability is broadly general. A livestream cannot, by itself, distinguish between a widely transferable learned policy and a system combining learned perception with task-specific motion primitives, state machines and exception rules. Nor does it reveal the full distribution of retries, misroutes, human interventions, damaged parcels or recovery times.
The claim that WALL-B is superior to conventional scripted automation therefore remains unproven from this test alone. The demo shows a capable integrated system; it does not establish how much of that capability transfers to a different robot body, warehouse layout or task.
Throughput and endurance are separate scoreboards
X Square’s stated success rate above 98% is a strong headline, but the public result does not provide a standardized failure taxonomy. It is unclear how the company counted retries, temporary misroutes, recoverable errors, human assistance or parcels that reached the end of the line incorrectly. 2
8
Figure’s headline strength is different. Three Figure 03 humanoids reportedly processed nearly 250,000 packages during 200 consecutive autonomous hours without hardware failures. 9
10
That endurance claim matters because a logistics customer needs more than a fast minute or hour. It needs predictable operation across shifts, charging, maintenance and changing failure conditions. At the same time, Figure’s 200-hour result does not establish that its parcel mix, orientation difficulty, effective yield or exception handling matched X Square’s one-hour setup.
Neither company has published enough normalized data to declare a universal winner.
Two different robotics strategies
Figure is making the humanoid bet: a mobile, human-shaped platform intended to work in spaces already designed around people and eventually move between tasks.
X Square’s public strategy is broader and more modular. It is using WALL-B across physical applications while developing robot bodies, data systems and industrial deployments. A high-throughput logistics cell can serve as a commercial beachhead even if the company’s longer-term goal is general-purpose embodied intelligence. 11
12
That strategy has an important implication: success in a fixed parcel station is evidence of a viable application, not proof that the same policy will work unchanged in a home, on a humanoid body or in an unstructured warehouse.
What the economics still do not tell us
Parcels per hour is only one part of an automation business case. A fixed dual-arm cell may be easier to justify when a site has high volume, engineered conveyor interfaces and a stable layout. Its actual return depends on acquisition or leasing cost, integration, maintenance, uptime, labor displaced and the number of parcels handled correctly—not gross speed alone.
A humanoid may eventually address more environments, but it must justify the additional hardware and control complexity. Its value depends on whether mobility and flexibility reduce the need for site redesign or enable tasks that fixed automation cannot economically cover.
X Square has said its WALL-B logistics system is operating on a live parcel-sorting line, and the company later extended the public challenge to a five-hour demonstration that processed 10,000 parcels in 5 hours, 14 minutes and 1 second. Those results are notable, but independent production-scale uptime and return-on-investment data remain unavailable. 13
Why X Square’s growth matters
The livestream also arrived as X Square was scaling rapidly. The Shenzhen company was founded in December 2023 and had completed four consecutive financing rounds through Series C by June 2026, when reports put its valuation above RMB20 billion, or about $2.8 billion. 14
15
Reported backers across its financing history include major Chinese technology and investment groups. That capital supports a dual-track approach: deploy specialized systems in commercially attractive settings while continuing to build foundation models, data infrastructure and more general-purpose robot bodies. 15
16
The funding does not validate the technical claims, but it helps explain why X Square can pursue both near-term logistics revenue and a longer-term general-robotics ambition.
What a fair robotics leaderboard would measure
The competing livestreams point toward a more useful style of robotics evaluation: unedited, time-stamped operational tests with published throughput and duration. That is better than judging systems from short promotional clips.
A genuinely comparable leaderboard would also need to standardize:
- parcel size, weight, material and distribution;
- feed rate and conveyor configuration;
- robot count and hardware configuration;
- successful-handling, retry, damage and misroute definitions;
- human-assistance and teleoperation logs;
- charging, maintenance and replacement procedures;
- uptime over days or weeks; and
- installed cost and energy use per correctly handled parcel.
Until those details are published, company-run livestreams should be treated as performance demonstrations rather than controlled head-to-head benchmarks. 8
The fair verdict
X Square Robot currently has the stronger public one-hour throughput result for a fixed parcel-induction cell: 1,816 parcels per hour at a reported success rate above 98%. Figure has the stronger public endurance result for humanoid autonomy: nearly 250,000 packages over 200 hours without a reported hardware failure.
The demonstrations reveal less about which company has built the universally better robot than about which design is optimized for which job. X Square’s result favors specialized industrial throughput; Figure’s favors mobile embodiment and sustained operation.
The decisive question remains unanswered: which system can process parcels correctly, affordably and independently in a real customer operation?
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "X Square Robot Beat Figure on Speed—but Not on the Benchmark That Matters"?
X Square Robot reported 1,816 parcels per hour with more than 98% success in an unedited one hour test—about 45% above Figure’s 1,247.8 package per hour average—but the demonstrations were not comparable head to head...
What are the key points to validate first?
X Square Robot reported 1,816 parcels per hour with more than 98% success in an unedited one hour test—about 45% above Figure’s 1,247.8 package per hour average—but the demonstrations were not comparable head to head... X Square used a stationary dual arm logistics cell; Figure used three bipedal humanoids over 200 hours.
What should I do next in practice?
The unresolved commercial metric is cost per correctly handled parcel in an independently observed customer operation, including interventions, retries, damage, uptime and maintenance.
Sources
humanoidsdaily.comNo Fingers, No Problem: X Square Robot's 1816-Parcel- ...
aol.comX Square Robot Demonstrates Embodied AI in Real-World Logistics ...
letsdatascience.comX Square Robot Reports 1,816 Parcels an Hour in Live Test
roboticsandautomationnews.comX Square Robot showcases full embodied AI portfolio at WRC 2026