No Fingers, No Problem: X Square Robot’s 1,816-Parcel-Per-Hour Livestream Challenges Figure

Key Takeaways

Hide

• X Square Robot completed a one-hour, unedited live logistics challenge today, successfully sorting 1,816 parcels with a success rate of over 98%.
• The dual-arm robotic system utilized purpose-built grippers instead of complex five-finger dexterous hands, prioritizing high-speed, cost-effective throughput for real-world deployment.
• The livestream format and barcode-orientation task directly mirror Figure AI's recent 8-hour unedited shift and subsequent 200-hour autonomous marathon.
• Founded in December 2023, the Shenzhen-based startup is backed by heavyweights like Xiaomi and Meituan, and is rapidly advancing its "General Embodied Foundation Model" architecture.

The high-stakes trend of live, unedited robotic endurance tests has officially crossed international borders. Today, Shenzhen-based embodied AI startup X Square Robot broadcasted its own unedited logistics challenge, pitting its dual-arm robotic system against a high-speed conveyor belt of varied parcels.

The objective was straightforward but demanding: sort 1,248 parcels in 60 minutes under real-world logistics conditions. By the time the hour was up, the system had comfortably shattered its own benchmark, achieving a final throughput of 1,816 randomly selected parcels per hour with a success rate exceeding 98%.

Grippers Over Hands

A central theme of X Square Robot’s demonstration was a deliberate step away from one of the humanoid industry's current obsessions: the five-finger dexterous hand.

Instead of outfitting its arms with watchmaker-grade, high-degree-of-freedom hands, the company opted for simple, purpose-built grippers. In statements released on X (formerly Twitter) surrounding the event, the company emphasized that its hardware choices are driven by commercial practicality rather than marketing flash.

"We often ask ourselves why we remain committed to a purpose-built gripper," the company stated. "The answer is simple: not for hype, and not for a demo, but for real-world deployment. Our goal is to deliver high efficiency at a lower cost."

Despite the mechanical simplicity of the end-effectors, the arms displayed an impressive level of dynamic control. The sorting task required the system to handle parcels of varying shapes, sizes, and materials—from rigid cardboard boxes to soft, irregular poly bags. The primary goal was to grasp and flip incoming packages so that their barcodes or QR codes faced upward, while pushing certain package types into alternative routing chutes.

The system's real-time adaptability was best highlighted during a moment when a package was precariously close to falling off the workstation. Rather than failing the sequence, the grippers successfully intervened, saving the parcel and sorting it correctly. The visual proof of this high-speed operation is captured in the company's broadcast.

The Figure AI and AGIBOT Parallels

For anyone following the intelligent automation sector over the last few months, the format of X Square's demonstration feels intimately familiar.

In May, Sunnyvale-based Figure AI captivated the industry by launching what was supposed to be an 8-hour unedited shift of its Figure 03 humanoids sorting packages. That stream quickly escalated into a viral 26-hour milestone, marched past the 64-hour mark, and ultimately concluded as a historic 200-hour marathon with zero hardware failures.

Similarly, Chinese developer AGIBOT recently shifted the public benchmark from the corporate lab to the factory floor. In late June, the company hosted a six-day global livestream broadcasting its humanoid fleet directly from Longcheer Technology’s active tablet mass-production line. During that unedited broadcast, multiple robots collaborated directly with human workers to manage a full quality-inspection section, proving their ability to handle the unpredictable edge cases and strict rhythms of a live industrial environment.

While Figure's demonstrations utilized a full bipedal platform autonomously swapping out to recharge, and AGIBOT proved its mettle in high-stress electronics mass-production, X Square’s livestream focused strictly on stationary dual-arm manipulation. However, the core technical hurdle across all these unedited broadcasts remains exactly the same: using vision-based AI foundation models to process raw camera pixels and generate real-time motor torque to manipulate unpredictable objects in a continuous, unscripted environment.

A $1 Billion War Chest for Embodied AI

While X Square Robot may not yet possess the household name recognition of some American humanoid developers, its rapid ascent and deep pockets make it a formidable player.

Founded in December 2023, the startup is driving what it calls a "General Embodied Foundation Model". The company has moved at a blistering pace, releasing its WALL-A and WALL-B models, building China's first large-scale industrial data acquisition facility, and open-sourcing its WALL-OSS ecosystem.

This speed has been fueled by massive capital injections. Between late 2024 and early 2026, X Square secured multiple financing rounds, including a 1 billion yuan Series A++ round in January 2026 and a subsequent Series B round in April 2026 backed by Xiaomi. The company's investor roster also includes major players like Meituan, Lightspeed China, and Legend Capital.

"Since the beginning of this year, we’ve worked through the entire loop—from collecting real-world data and training the model to continuously testing and refining the system," X Square noted following the stream.

X Square Robot has confirmed it will be showcasing its complete logistics sorting line live at the upcoming World Robot Conference (WRC) in Beijing. As the battle for embodied intelligence moves from scripted, highly edited videos to raw, unedited livestreams, the pressure is mounting on hardware developers globally to prove their systems can actually survive the dull, dirty, and demanding realities of a modern logistics line.