AGIBOT Launches Genie Studio 2.0 to Bring Robots into Real Workplaces

【SHANGHAI, China, October 26, 2026】AGIBOT today launched Genie Studio 2.0, an integrated platform for developing, deploying and operating embodied AI applications. The platform connects robot models, task workflows and on-site systems to help turn model capabilities into reliable, repeatable operations in real workplaces. It supports initial deployment and validation in a new task environment in as little as seven days.

Genie Studio 2.0 brings these functions together in three modules: Flywheel for model development and iteration, Agent for modular task orchestration, and Monitor for fleet operations. Together, they cover data collection, model training, evaluation, inference, task orchestration and fleet management.

Flywheel adapts robot skills to specific tasks

Flywheel helps robots adapt to the processes, materials and precision requirements of a particular workplace. It connects data collection, model training, simulation testing and deployment on physical robots, with failure and anomaly data fed back into the next training cycle.

Flywheel connects data management, model training, simulation evaluation and deployment on physical robots.

Data collection. Flywheel manages data from physical robots, simulations and open-source datasets in one place. Data from physical robots captures visual inputs, motion trajectories and execution status. Teams can use Genie Sim to generate simulation data for situations that are costly to capture, occur infrequently or are risky to test on hardware. Open-source datasets such as AGIBOT WORLD extend coverage across tasks and environments.

Model training. Customers can train visual perception models, vision-language-action (VLA) models and reinforcement learning models for physical robots. The platform also provides access to AGIBOT’s GO-2 embodied foundation model, along with tools for fine-tuning, experiment management and compute resource scheduling. In real-world project validation, AGIBOT’s in-house reinforcement learning training framework achieved convergence within hours, with the relevant tasks reaching a 99% success rate.

Simulation testing. Trained models can be tested before deployment on physical robots. Genie Sim Benchmark covers four evaluation dimensions and 42 task categories, with integration taking five minutes and evaluation completed in 30 minutes. The difference between simulation results and performance on physical robots is less than 10%, helping teams identify potential problems before deployment.

Deployment and feedback. Once robots are operating, failure and anomaly data returns to Flywheel for further training. This feedback loop allows teams to address problems encountered on site and reuse the resulting improvements in subsequent deployments.

Agent combines robot skills into complete workflows

A single skill is only one part of a complete job. Loading and unloading a machine tool, for example, can involve dozens of steps, including receiving production instructions, navigating to collect material, identifying and grasping a workpiece, transferring it, interacting with equipment and recovering from errors.

Agent lets teams assemble these steps through a visual drag-and-drop interface. Navigation, visual localization, model inference and decision logic can be combined into complete workflows, including calls to VLA and reinforcement learning models. Development tools allow teams to add custom business logic and connect to third-party systems.

 Agent combines visual workflow orchestration with simulation to support task development and testing.

Before deployment, teams can test applications in simulation to check paths, potential collisions, robot poses and workflow logic. Validated workflows, skill nodes and parameter configurations can then be packaged as templates for reuse across projects.

Monitor manages robot fleets in operation

Monitor provides a centralized view of each robot’s status, task progress and operating data. Operators can group robots and schedule tasks across a fleet, reducing the need to manage each device individually.

When a fault occurs, Monitor records it, issues an alert and supports remote intervention and online debugging. Operators can identify the affected robot and workflow step without logging into each device separately or searching through individual logs.

Monitor provides a centralized dashboard for fleet status, task management and remote intervention.

Operating data, anomaly records and resolution details can be fed back into Agent and Flywheel to refine workflows and update models. This connects day-to-day fleet operations with continued improvements in robot performance.

Validation in manufacturing and rail transit

Genie Studio 2.0 has been used in real-world manufacturing and rail transit projects, covering deployment validation, precision assembly and fleet operations.

CNC machine tending in Changzhou

A project involving six CNC machines and collaboration between robots and automated guided vehicles (AGVs) progressed from requirements definition to production-line validation in 14 days. Flywheel supported data and model preparation, while Agent orchestrated navigation, material pickup, machine loading and unloading, placement and exception handling. Routes, motion sequences and potential collisions were checked in simulation before testing on physical robots.

During on-site integration, the robots were connected to programmable logic controllers (PLCs), manufacturing execution systems (MES), vision systems and fixtures. Monitor collected device status and task information. Following continuous testing, the workflows, skill nodes, parameters and interface configurations were consolidated into templates for reuse on similar production lines.

BMS controller assembly at PIA Automation

On PIA Automation’s pilot validation line, three AGIBOT G2 robots worked together to load and assemble components for automotive battery management system (BMS) controllers, pack finished products into boxes and transfer material bins. The two companies used Genie Studio 2.0 to orchestrate tasks and validate workflows in simulation before execution on the line.

Working with highly reflective metal parts and placement accuracy requirements of less than 1 mm, the precision loading, unloading and assembly tasks consistently achieved success rates above 99.9%, with cycle times of 13 seconds or less per operation. The line also completed validation of stable, round-the-clock production, operating 24 hours a day, seven days a week.

Tablet loading and unloading at Longcheer Technology

At Longcheer Technology’s tablet loading and unloading line, Genie Studio 2.0 centrally manages multiple robots and tracks operating hours, task counts, success rates, cycle times and faults. When an exception occurs, operators can see which workflow step the robot has reached. The data is also sent to the cloud for analysis and further optimization.

More than 60 devices across Guangzhou Metro stations

Genie Studio 2.0 has supported the deployment of more than 60 devices for services at Guangzhou Metro stations. A centralized dashboard displays device status, task success rates and mean time between failures across stations. Fault and warning notifications, remote connections and quick commands help reduce reliance on on-site inspections.

Deploying embodied AI still requires on-site commissioning, system integration, sustained operations and the ability to replicate applications across environments. Genie Studio 2.0 connects data, models, robot skills and on-site systems into applications that can be deployed, operated and continuously improved.

As teams build up reusable skills, workflows and templates, AGIBOT aims to reduce the need for custom development from scratch, accelerate application delivery, improve long-term reliability and lower the overall cost of using robots.

To explore Genie Studio 2.0, please visit https://genie.agibot.com/geniestudio.

For more information, please visit AGIBOT.com and follow AGIBOT on:
X: https://x.com/AGIBOTofficial
LinkedIn: https://www.linkedin.com/company/agibot/
YouTube: https://www.youtube.com/@AGIBOTofficial
Instagram: https://www.instagram.com/agibotofficial/
TikTok: https://www.tiktok.com/@agibotofficial
Facebook: https://www.facebook.com/AGIBOTofficial/

About AGIBOT
AGIBOT is a pioneer in the global general-purpose AI robotics industry, with its core technology focused on embodied AI. AGIBOT's "Three Intelligences in One" architecture integrates Locomotion Intelligence, Interaction Intelligence, and Manipulation Intelligence into a unified embodied system. Its portfolio spans humanoid robots, quadrupeds, dexterous systems, and commercial cleaning solutions. In September 2026, AGIBOT announced that its 20,000th robot had rolled off the production line.