Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

Matrix Robotics demonstrates MATRIX-3 in first Hong Kong retail scenario

Matrix Robotics demonstrated MATRIX-3 at Hong Kong Fintech Week on 2-3 November in its first real-world retail scenario in Hong Kong, with a starting price of 580,000 yuan per unit.

AI & modelsAnalysisRachel NwosuPublished: 5 November 20263 min readSources 24
Matrix Robotics demonstrates MATRIX-3 in first Hong Kong retail scenario

Visitors scanned a QR code to give MATRIX-3 a task in an unmanned retail setting. Matrix Robotics said the robot then completed the chain without preset fixed scripts: recognising the target, grasping a small item, walking and turning, and handing the item to a person.

The demonstration took place at Hong Kong Fintech Week x StartmeupHK, held from 2 to 3 November at the Hong Kong Convention and Exhibition Centre. Matrix said the event hosted a robotics and embodied intelligence exhibition area for the first time, and that it showed MATRIX-3 alongside several other Chinese embodied AI companies.

What Matrix disclosed

Matrix published commercial details with the Hong Kong demo. MATRIX-3 stands 1.70 metres tall and weighs 65 kilograms. Its rope-driven hand has 27 degrees of freedom, with fingertip sensing accurate to 0.1 newtons. The company says the robot can operate continuously for up to four hours at a time, and that its price starts at 580,000 yuan per unit.

Matrix said that configuration and price give retail and commercial buyers a calculable economic basis to move the robot from trial use to regular use. The Hong Kong announcement did not name paying customers or give deployment volumes.

Third public showcase in five months

The Hong Kong appearance is Matrix's third public showcase in five months. In June, it was the sole humanoid robotics service provider at the Summer Davos in Dalian, where MATRIX-3 worked as an AI barista for attendees. At WAIC 2026 in Shanghai, which ran from 17 to 20 July, the company showed four hardware-identical MATRIX-3 units across four live demonstration zones.

WAIC was also where Matrix announced the latest upgrade to WAVE, its physical world foundation model, and launched the MATRIX Global Partner Program for regional market expansion. It gave updates on an existing RAAS ecosystem partner program. The announcement did not give territories, targets or signing dates for the global program.

WAVE uses a brain-cerebellum architecture. The company says the brain handles scene understanding, reasoning and task planning, while the cerebellum generates full-body motion. Matrix trains the model on real and simulated interaction data and feeds operating data from robots in the field back into training.

What the demo leaves open

The Hong Kong demo shows a robot taking an instruction from a stranger's phone in a noisy hall. It does not show how often staff had to intervene, the robot's success rate or its cycle time. Those measures would let a retailer compare the robot with fixed automation.

Matrix says competition in humanoid robotics is shifting from demonstration capability to task delivery in real-world, open scenarios. Its Hong Kong disclosure supports that claim with a price and a runtime. The missing piece is a named customer or a deployment count that would show the robot working outside a showcase.

Comments 0

Sources

24
  1. 01Matrix Robotics MATRIX-3 Makes Its Debut At Hong Kong Fintech Week, Completing Its First Real-World Retail Scenario Demonstration In Hong KongEN
  2. 02Matrix Robotics Debuts At Dalian Summer Davos As Sole Humanoid Robotics Service Provider, With MATRIX-3 Serving Global Attendees As An AI BaristaEN
  3. 03Matrix Robotics Showcases MATRIX-3 at WAIC 2026, Announces WAVE Physical World Foundation Model Upgrade and Launches Global Partner ProgramEN
  4. 04Matrix Robotics Showcases MATRIX-3 at WAIC 2026, Announces WAVE Physical World Foundation Model Upgrade and Launches Global Partner ProgramEN
  5. 05Matrix Robotics showcases MATRIX-3, launches partner programEN
  6. 06Churchill Capital XI shareholders approve $2.5B Agility Robotics humanoid SPACEN
  7. 07Xpeng brings G9L to Hong Kong with extended-range powertrain after European launchEN
  8. 08Tokenised gold bar to advance Hong Kong’s ambitions as digital, metal trading hubEN
  9. 09Estonia approves 24/7 robotic pharmacies, paving the way for a European firstEN
  10. 10How Hong Kong’s fintech strategy is progressing – and where growth still lagsEN
  11. 11Has Hong Kong’s IPO boom hit a wall – or is a year-end rush still to come?EN
  12. 12Hong Kong’s Deadly 2025 Fire: Accountability on Lee’s TermsEN
  13. 13Securing a Global Giant: HKSTP Lands Merck to Power Hong Kong’s I&T AmbitionsEN
  14. 14Xiaomi’s humanoid robots work a full 8-hour shift in its EV factoryEN
  15. 15Robotics Startup Radar 2026EN
  16. 16How GlobalLogic helped Haystack Robotics build a UV disinfection robotEN
  17. 17China October NEV retail sales fall 10% to 1.148 million, preliminary CPCA data showsEN
  18. 18Gentherm designs ThermAffyx to secure, heat patients during robotic surgeriesEN
  19. 19Top 10 robotics developments of October 2026EN
  20. 2028 Zero-Day Findings in Less Than Two Days: What AI-Speed Discovery Means for RoboticsEN
  21. 21Solving the Interconnect Challenge in Modern Robotics: How 3M Is Engineering for Motion, Density, and UptimeEN
  22. 22Unusual robotic animals can prove their utilityEN
  23. 23In a global first, Hong Kong opens secondary trading for tokenised public fundsEN
  24. 24ByteDance accelerates robotics commercialisation with new recruitment driveEN

All figures and quotations in this text come from the sources listed below.

Content prepared by the editorial team with AI assistance.

Rachel Nwosu

Rachel Nwosu

AI, models and technology

Rachel Nwosu covers AI, models and technology for FLASH24, working from public model documentation, benchmark releases and repository histories rather than press summaries, and she skips announcements that arrive without reproducible numbers. She checks training-data claims against dataset cards and reruns reported metrics where code is available. She spends much of her week interviewing researchers and engineers, tracking model launch calendars, and comparing vendor benchmarks with independent evaluations. Outside the desk she runs 3D printers, restores old computers, and tests how models learn from internet junk. She does not publish benchmark figures she cannot trace to a source.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

Comments are public. We do not publish abuse, spam or advertising.