AgiBot tests 24-hour convenience store operation with six Genie G2 robots, completes 1,730 tasks

AgiBot Innovation (Shanghai) Technology Co., Ltd. announced on November 9, 2026, that six Genie G2 robots had operated continuously for 24 hours at a convenience store and completed a total of 1,730 tasks. The test ended at 10 a.m. on November 3, and the company said four robots had been “made permanent.”

AgiBot’s announcement gives the scale of a field test in which six Genie G2 robots operated for 24 hours at a convenience store and completed 1,730 tasks. Assessing robot operations in stores requires checking task counts alongside completion criteria, human intervention and operating status.

The reported figures — six robots, 24 hours and 1,730 tasks — indicate the scale of an operational test run for a set period in a convenience store environment, not performance on individual tasks. AgiBot also said four robots had been “made permanent,” but did not explain the selection criteria or what that status specifically entails. The 1,730 tasks therefore cannot be interpreted as successful or autonomously completed tasks, nor can the four robots be considered capable of unstaffed operation. The key measure for evaluating this announcement is not the task count alone, but the metrics used to assess continuous operation in a real-world environment. The test can be seen as the post-test results for the 24-hour trial using six G2 robots, announced in AgiBot to unveil seven new products at AI WAVE; 24-hour G2 convenience store test to begin Nov. 2. The earlier announcement outlined the plan; this update reported operating time and the total number of tasks. However, the announcement does not explain how the “on-the-job test” mentioned in the earlier article relates to the “made permanent” status, or how continued operation and task completion were determined. A notable aspect of the industry context is that AgiBot presented its field test alongside a full-stack vision at the same AI WAVE event, covering datasets, robots, simulation and task platforms, an agent framework, a foundation model and collective learning. AgiBot unveils embodied agent framework Genie Master, tests autonomous improvement in a real store and AgiBot, Tsinghua University and others announce CEL, test collective learning with 96 robots covered the agent framework and collective learning, respectively. But the announcement does not say whether Genie Master, CEL or other components were used in the convenience store test, how task data were recorded and evaluated, or whether the data fed back into model updates. The existence of these technologies should be distinguished from their use and impact in the field test. The next details to check are the categories of the 1,730 tasks and the criteria for completion, records of failures and human intervention, the operating status of each of the six robots, and how “made permanent” was determined. If provided, these details would make it possible to compare results from a single livestreamed event with metrics for reproducible store operations. Based on the figures released so far, it is not possible to assess the range of tasks, store conditions or specific links to the seven technical achievements.

AgiBot reported the test as the finale of AI WAVE. It said the livestream drew more than 2 million viewers across all online platforms. The seven items AgiBot presented at AI WAVE were Physical Agent, the fourth installment of the AGIBOT WORLD 2026 dataset; the heavy-duty robot G2 Max; the simulation platform Genie Sim; the task platform Genie Studio 2.0; the embodied agent framework Genie Master; the GO-3 Preview foundation model; and the collective experience learning paradigm (CEL). AgiBot described them as part of a full-stack AI effort connecting data, models, robots and field operations.