AgiBot G2 Max humanoid robots take inbound pallet work at JD Logistics' Zhi Lang warehouse

Humanoid robots just showed up at the front door of one of China's busiest automated warehouses, not on a dance floor. AgiBot unveiled its safety-certified Elf G2 Max heavy embodied robot on October 26, 2026, and JD Logistics said the same week the unit is running inbound work inside its Zhi Lang "goods-to-person" site. ITBear's October 27 report describes pallet breaking, stacking, box moves, and standard container handling at the start of the warehouse flow, with AgiBot claiming more than 99 percent task success, 80 boxes depalletized per hour, and 75 boxes palletized per hour. For buyers, sellers, and operators trading humanoid robots on Buy My Robots, the headline is simple: a volume Chinese maker paired a biped-class heavy loader with a retailer-owned mega-DC during peak-season pressure, which is a different diligence packet than Guinness choreography or metro guidance pilots.

What AgiBot shipped with the G2 Max launch

AgiBot positions G2 Max as an industrial humanoid built for depalletizing, material transport, and line feeding rather than guest-facing animation. 42HOW's October 26 launch recap lists seven-degree-of-freedom force-controlled arms rated at 20 kg per arm, 40 kg dual-arm nominal load, and 50 kg peak, with repeat positioning within 0.5 mm. The vertical work envelope runs from 0.1 m to 2.5 m with about 1.2 m horizontal reach, and a waist pitch joint lets the torso lean so pallets can sit against a wall without the 1.5 m clearance many AMR stacks need behind a rack.

Power and safety read like factory procurement, not lab demo. AgiBot advertises dual hot-swappable batteries plus autonomous return-to-charge for continuous shifts, and wraps the chassis in 2D/3D lidar, ultrasonics, bump strips, and a physical e-stop. Software routes through Genie Studio 2.0 for task orchestration, simulation checks, and fleet ops, which matters when you resell a unit that was configured inside someone else's WMS rules.

The company also named non-JD sites: mixed FMCG lines doing depalletize, open-box feed, film cover, and empty-tote recovery, plus AgiBot's own Qiyuan T1 assembly line where managers told trade press the robot's feed cadence can track a new humanoid every 2.5 minutes. Those are parallel proof points, but the JD Zhi Lang placement is the one moving on October 27 because it sits inside a named third-party logistics operator during shopping festival season.

Why JD put a humanoid on the inbound dock

Zhi Lang is JD Logistics' in-house goods-to-person stack. Sina Finance, citing JD's internal channel post, says the building uses roughly 12 meters of vertical storage, claims about six times the storage density of a conventional warehouse, and runs hundreds of robots at once with outbound picking efficiency up about threefold. Inbound is the choke point: if depalletizing or tote stacking stalls, every downstream shuttle and picker inherits the delay.

JD's stated reason for hosting G2 Max is blunt. Real supply-chain volume is the test bench for embodied AI, not a staged booth. The operator already markets a "Wolf Pack" portfolio spanning storage, picking, transport, and delivery bots coordinated by its Chao Nao large model. Adding a biped-class loader at the dock is an experiment in whether humanoid morphology buys flexibility on mixed carton and tote shapes without ripping out the existing wolf herd.

AgiBot's published hourly box counts are the numbers a secondary-market buyer will quote back without context. Treat them as vendor-reported trial metrics inside one facility during a launch window until JD or AgiBot publishes shift logs, intervention counts, and damage rates with third-party review.

How this differs from our October 25 AgiBot Guinness post

Our October 25 Jinan record piece tracked 110 Expedition A3 units in a certified swarm dance with LinkBots scheduling and UWB spacing. Today's story is Elf G2 Max on industrial inbound pallets inside JD's Zhi Lang warehouse, with load, precision, and boxes-per-hour claims tied to e-commerce logistics. Same parent company, different SKU class, different buyer question: fleet choreography for venues versus depalletize throughput at a goods-to-person hub.

What marketplace users should verify before paying a "JD pilot" premium

A warehouse headline does not travel with the robot unless paperwork says it does.

Confirm model series (G2 Max versus Expedition or earlier G2 variants), firmware build, and whether Genie Studio project files export cleanly to your site.
Ask for task logs on depalletize and palletize cycles, not only success-rate percentages in a press release.
Map safety certification geography. AgiBot calls G2 Max safety-rated for industrial use; your insurer and local regulator may still treat Chinese embodied platforms differently than Digit 5 or Atlas deployments covered in Western plant announcements.
Compare economics to non-biped automation already on the dock. If a palletizer plus conveyor hits the same 80-box hour with lower intervention, the humanoid is a flexibility bet, not a capex discount.
If you list on Buy My Robots, disclose whether the unit ever ran inside JD's fence or only on AgiBot's Qiyuan line, and whether JD's data-retention rules block transfer.

Western import paths remain a separate gate from performance video. FCC and covered-list policy posts in our archive still apply to many Chinese humanoid SKUs regardless of JD's domestic scale.

What to watch after October 27

Three signals will tell you if G2 Max becomes a comps chart or stays launch theater. First, whether JD expands beyond inbound Zhi Lang to other wolf-family sites with named box counts per site. Second, whether AgiBot's October 29 Innovation Day demo adds third-party timing on battery hot-swap and mean time between human interventions. Third, whether competing logistics giants answer with their own biped inbound pilots or double down on dedicated depalletizing arms, which would reprice used G2 units aimed at warehouse resellers.

Until then, AgiBot put a heavy, safety-framed humanoid on a real JD inbound line with quantified box-throughput claims while peak season stress-tests the dock. Price the task evidence, not the brand halo.