AGIBOT to unveil seven new products at AI WAVE; G2 to undergo 24-hour convenience-store test starting Nov. 2

AGIBOT Innovation (Shanghai) Technology Co., Ltd. announced Oct. 22 that it would unveil seven new items in succession through its “AI WAVE” from Oct. 22 to Oct. 30. They are an embodied reasoning data set, Genie Studio 2.0, G2 Max, Genie Master, Genie Sim, a foundation model and a prototype validation of collective experience learning (CEL). The company also said it would deploy six G2 robots at a convenience store from Nov. 2 to 3, livestreaming their work for 24 hours and testing four shifts from 10 a.m. to 10 a.m. the following day. The announcement foregrounds the company’s view that large-scale robot deployment requires full-stack coordination.

What AGIBOT outlined on Oct. 22 was not simply a promotional push for robot hardware, but a plan to link data collection, virtual testing and field operations in one workflow. The 24-hour test with six robots at a convenience store will offer a way to assess whether the announced technologies work in a real environment. For businesses considering deployment, execution of the operating model matters more than standalone performance.

The new element is not just a product announcement, but the presentation of seven layers as one package: a data set, an operations platform, a robot, an agent framework, a simulation platform, a foundation model and collective experience learning. At the center is not a contest over the performance of individual robots, but the “full-stack coordination” repeatedly emphasized in the Oct. 22 announcement. By bringing together products, a development platform and a testing environment, as with G2 Max on Oct. 26 and Genie Sim on Oct. 28, the focus appears to be preparing the upstream and downstream processes needed to keep physical AI operating, rather than launching a new hardware model. In a related article, [AGIBOT Innovation features Expedition A3, showcasing martial arts, table tennis and ballroom dancing with mass-produced robots](/news/satoru-moto-soushin-ensei-a3-wo-tokushuu-ryousanki-de-bujuts-16d6f0), we described how the applications of mass-produced robots were expanding into sports and cultural-tourism uses. This time, the company says it will put six G2 robots in a convenience store, an everyday operating environment, and test them across shifts over 24 hours as they stock shelves, replenish goods, organize merchandise, prepare hot food, collect trash, make coffee and greet customers. In other words, the emphasis has shifted from “how to showcase mass-produced robots” to “how to demonstrate continuous operation.” This suggests the product pitch is moving beyond exhibitions and one-off demonstrations toward testing that includes operational processes and maintenance. In terms of industry structure, the company’s simultaneous announcement of a data set, simulation, a foundation model and an operations platform suggests that competition is shifting beyond robot specifications to an end-to-end cycle that includes acquiring training data, reproducing conditions in virtual environments and retraining after field deployment. In particular, the references to the “first batch of 150 hours of real-robot data” and “more than 95% real-to-simulation consistency” show the importance of collecting field data and reusing it in virtual environments. However, the robots’ actual operating stability, the types of tasks handled by the six G2 units and their respective share of the work, task-specific failure rates and frequency of human intervention remain unclear. The next points to watch are which of the seven items will lead commercialization, what specific load conditions G2 Max’s “safety-standard heavy load” refers to, and whether the 24-hour convenience-store test will remain a one-off event or lead to ongoing operations. Also in focus are how the 150 hours of real-robot data will be expanded and how many robots will be involved in further prototype validation of CEL.

The Oct. 22 announcement listed a plan to unveil the items in succession: on Oct. 22, the first 150 hours of real-robot data for “AGIBOT WORLD 2026, Phase 4: Embodied Reasoning (Physical Reasoning)” data set; on Oct. 23, “Genie Studio 2.0”; on Oct. 26, “G2 Max, a safety-standard heavy-load embodied industrial robot”; on Oct. 27, “Genie Master”; on Oct. 28, “Genie Sim”; on Oct. 29, a foundation model; and on Oct. 30, prototype validation of collective experience learning (CEL). The company said real-to-simulation consistency for Genie Sim exceeds 95%. At the convenience store chosen for the field test, six G2 robots are expected to stock shelves, replenish goods, organize merchandise, prepare hot food, collect trash, make coffee and greet customers. The Oct. 22 announcement described this as “real work.” It said robots cannot operate in the physical world without data, hardware, tools, systems and models, and presented the successive announcements and field test as parts of one plan.

For Japan’s retail, food service and logistics sectors, deployment decisions are often shaped less by the robot itself than by the collection of field data, virtual testing and the design of long-duration operations. AGIBOT’s announcement suggests that the competitive focus is shifting from hardware alone to data and operational infrastructure. Japanese companies, too, are likely to be judged not just on their robots, but on systems that put real-world operating logs to use.