G2

Source-backed capabilities

• Operating environments: Longcheer tablet production line
• Control modes: Autonomous
• Machine tending

  Commercial deployment - 2026-06-05 - Load and unload tablets at Longcheer MMIT testing stations. - environment: Longcheer tablet production line

  Source
• Object sorting and classification

  Commercial deployment - 2026-06-05 - Separate finished and defective tablets after the MMIT testing workflow. - environment: Longcheer tablet production line

  Source

• Operating Milestone

  More than 64 hours - 2026-06-29 - Longcheer - production-line validation - AgiBot reported more than 64 operating hours during a six-day Longcheer validation.

  Source
• Operating Milestone

  64,828 tasks - 2026-06-29 - Longcheer - production-line validation - AgiBot reported 64,828 completed production-line tasks during the validation.

  Source
• Operating Metric

  99.99% - 2026-06-29 - Longcheer - production-line validation - AgiBot reported 99.99 percent task success during the validation.

  Source
• Operating Milestone

  up to 310 UPH - 2026-04-14 - Longcheer Technology - consumer-electronics precision manufacturing - The AgiBot G2 Longcheer deployment was reported with line performance of up to 310 units per hour.

  Source
• Operating Metric

  99%+ success rate - 2026-04-14 - Longcheer Technology - consumer-electronics precision manufacturing - The AgiBot G2 Longcheer deployment was reported with a 99%+ task success rate.

  Source
• Operating Milestone

  roughly 3,000 units per shift - 2026-04-14 - Longcheer Technology - consumer-electronics precision manufacturing - The AgiBot G2 Longcheer deployment was reported to process roughly 3,000 units per shift.

  Source

• Customer Site

  1 Longcheer production-line site - 2026-04-14 - Longcheer Technology - consumer-electronics precision manufacturing - AgiBot G2 robots were reported as deployed at Longcheer Technology MMIT stations in a live consumer-electronics production environment.

  Source

Supply / Components

• Agibot G2

  Jetson Thor Edge AI Module - Compute & Control - NVIDIA - Agibot states G2 will fully integrate NVIDIA Jetson Thor - source: agibot.com

  Source
• Jetson Thor

  Compute & Control

• WorldOracle: Benchmarking Action-Conditioned World Models against Real-World Executions

  2026-10-13 - Research Publication - WorldOracle compares action-conditioned world-model rollouts with paired physical executions under the same starting observation and low-level actions. Its 335-example evaluation spans AgiBot G2, YAM and human-hand embodiments, including successful, failed and long-duration interactions.

  Source
• AgiBot World Challenge 2026 moves embodied AI evaluation onto G2 robots

  2026-06-05 - Research Publication - AgiBot reported that the World Challenge 2026 advanced from simulation to real-robot testing, with finalists completing tasks on AgiBot G2 robots. The source links G2 to embodied-AI benchmarks covering reasoning-to-action, world modeling, navigation, and whole-body manipulation.

  Source
• AgiBot World 2026: Real-World Embodied Intelligence Dataset

  2026-04-07 - Research Publication - Embodied-AI training needs structured data from variable real environments rather than only controlled laboratory demonstrations.

  Source
• Genie Sim 3.0: A High-Fidelity Comprehensive Simulation Platform for Humanoid Robot

  2026-01-05 - Research Publication - Genie Sim combines scene generation, synthetic manipulation data and automated policy evaluation. The paper describes more than 10,000 hours across 200 tasks; four-task physical transfer experiments use AgiBot G1, while G1 and G2 supply simulated data and benchmarks.

  Source
• FiveAges robot_usds simulation asset library

  Research Publication - A shared library of robot USD assets for ROS 2 control and Isaac Sim development, with configurable robot and end-effector models.

  Source

• AgiBot introduces WorldOracle benchmark for action-conditioned world models

  2026-10-13 - Research Publication - The 335-example evaluation spans AgiBot G2, YAM and human-hand embodiments with successful, failed and extended interactions. Seven world models are tested across rigid, deformable, fluid and articulated-object interactions. Metrics assess physical-state correctness, spatial alignment, interaction-area fidelity and coherence across synchronized views.

  Source
• AgiBot G2 logs factory task and machining deployments through Puzhi Robotics

  2026-07-22 - Deployment - The G2 completed an eight-hour, 2,283-task zero-error test at a domestic 3C electronics company. The robot is also working on an automotive safety-belt aluminium-part tapping process, improving the fastest cycle time to 12.9 seconds. Puzhi Robotics said its Wuxi subsidiary’s first G2 batch rolled off in June, with August capacity expected to reach 300 to 400 units.

  Source
• AgiBot deploys G2 robots on Longcheer tablet line

  2026-04-14 - Deployment - The G2 workflow loads and unloads tablets, places devices in test fixtures and sorts finished or defective units without custom tooling. Reported operations took 19–20 seconds with more than 99% success; expansion to 100 robots by the third quarter remained a plan.

  Source
• AgiBot releases AGIBOT WORLD 2026 dataset

  2026-04-07 - Research Publication - The G2-based collection combines Swift Picker grippers and OmniHand configurations with RGB-D, touch, lidar, IMU, joint-state and force information. Task descriptions, action steps, atomic skills and object labels include error-recovery trajectories; five research themes were planned for phased release.

  Source
• AgiBot introduces its robot portfolio to the US market at CES

  2026-01-05 - Product Launch - The company positioned A2 for guided presentations and showrooms, X2 for entertainment and education, and G2 for industrial manipulation. The Powered by AGIBOT programme offered customization of robot appearance, hardware, software and algorithms for partners entering local markets.

  Source

Market context

AgiBot G2 is a wheeled mobile manipulation robot for factory and commercial workflows that need mobile reach, force-controlled handling and embodied task execution.

• Target environment: Factory floors, commercial service environments and structured indoor workspaces.
• Workflow context: Mobile manipulation, object handling, force-controlled task execution and indoor workflow automation.
• Deployment model: Finished robot platform deployed for mobile manipulation tasks under controlled operating conditions.
• Commercial maturity: Public mobile manipulation product for factory and commercial task automation.
• Adoption constraints: Requires task setup, safety validation, site integration and manipulation reliability for customer workflows.
• Market position: Mobile manipulation platform within AgiBot's embodied robotics range for factory and commercial task automation.
• Adjacent products: Indoor mobile manipulator for object handling, force-controlled work and workflow automation.

Comparable Systems

• Elixis-W

  Addverb - Mobile Manipulator

• Galbot S1

  Galbot - Semi Humanoid

• TORA-ONE

  PaXini - Semi Humanoid

• Vega 1 Pro

  Dexmate - Semi Humanoid

• Casbot W1

  CASBOT - Mobile Manipulator

• Duatic Alpha

  Duatic - Semi Humanoid