AGIBOT Innovation Expertise Co. right this moment launched Genie Sim 3.0, a next-generation robotic simulation platform powered by NVIDIA Isaac Sim, at CES 2026. The corporate mentioned it delivers a unified, open simulation workflow that brings collectively digital asset technology, scene generalization, information assortment, automated analysis, and physics-based simulation in a single toolchain.
AGIBOT mentioned it designed Genie Sim’s core module, Genie Sim Benchmark, as a standardized analysis system to ascertain an correct and authoritative benchmarks for embodied intelligence.
Genie Sim 3.0 attracts from greater than 10,000 hours of artificial datasets, together with real-world robotic operation eventualities. The system integrates 3D reconstruction with visible technology to create a high-fidelity simulation setting. It additionally makes use of giant language mannequin (LLM) know-how to generate scenes and analysis metrics.
The analysis system covers greater than 200 duties throughout over 100,000 eventualities to construct a complete functionality profile for fashions. AGIBOT mentioned the platform can speed up mannequin growth, scale back reliance on bodily {hardware}, and empower innovation in embodied intelligence.
The Shanghai, China-based firm was established in February 2023. AGIBOT gives a spread of humanoid robots and cell robots for cleansing and repair settings.
In December, AGIBOT launched its up to date humanoid, AGIBOT A2, in December. Later that month, the corporate rolled out its 5,000th mass-produced humanoid robotic at its manufacturing facility.
Genie Sim 3.0 offers a full-stack pipeline for coaching and analysis
AGIBOT mentioned its built-in applied sciences allow it to ship real-time, high-fidelity, and high-precision simulation environments.
The system captures real-world environments with Skyland Innovation’s MetaCam handheld 3D laser scanner. This scanner combines high-resolution RGB photos, 360° lidar level clouds, and centimeter-level RTK (real-time kinematic) positioning. Interactable objects might be transformed into simulation-ready belongings from a single 60-second orbital video, considerably accelerating scene building, mentioned the corporate.
A defining characteristic of Genie Sim 3.0 is its LLM-driven natural-language scene technology, AGIBOT claimed. Customers can describe environments conversationally, and the system robotically produces structured scenes, visible previews, and 1000’s of semantic variations with out handbook logic coding. Imaginative and prescient-language fashions additional refine and tune these scenes to satisfy specification-level wants, enabling speedy adaptation and powerful mannequin generalization, it claimed.
As well as, Genie Sim 3.0 integrates actual industrial-scene datasets into each coaching and analysis pipelines. This method considerably shortens algorithm-validation cycles and reduces dependence on {hardware} infrastructure.
AGIBOT offers over 10,000 hours of artificial information
AGIBOT has additionally launched over 10,000 hours of open-source artificial information protecting greater than 200 duties with multi-sensor modalities reminiscent of RGB-D, stereo imaginative and prescient, and whole-body kinematics. An clever data-collection toolkit helps each low-latency teleoperation and automatic process programming. With auto-annotation and a restoration mechanism that resumes assortment after process failures, it will possibly scale back the price and time of dataset manufacturing.
For analysis, Genie Sim 3.0 offers greater than 100,000 simulation eventualities, which AGIBOT mentioned strikes past single-metric benchmarks to assemble full-spectrum functionality profiles for embodied fashions. Utilizing LLMs, the platform can auto-generate executable analysis workflows throughout numerous semantic, spatial-reasoning, and manipulation dimensions. The corporate mentioned it outlines mannequin strengths, limitations, and optimization pathways.
All simulation belongings, datasets, and automatic analysis supply code are totally open-source. Builders and researchers can entry AGIBOT Genie Sim on GitHub.





