AGIBOT WORLD CHALLENGE Open Session Simulation Evaluation Platform Opens for Internal Testing: Building a 7×24 Open Embodied Intelligence Evaluation Platform

On June 28, AGIBOT officially announced that the AGIBOT WORLD CHALLENGE: Open Session simulation evaluation platform has opened for internal testing. Built on the simulation evaluation benchmark Genie Sim Benchmark, the platform draws on the mature experience of two AGIBOT WORLD CHALLENGE competitions at IROS 2025 and ICRA 2026, and aims to provide global research teams and algorithm developers with an open, continuous, and high-confidence model evaluation platform.

The evaluation platform was jointly initiated by AGIBOT with Hugging Face, NVIDIA, and Tsinghua University, and invites renowned academic teams from around the world to participate, covering industry, academia, and research perspectives. Its launch marks AGIBOT's successful transformation of a top-tier competition's evaluation system into regular operations. Leveraging a high-fidelity simulation environment, simulation evaluation results can serve as a reliable reference for real-robot performance, building a simulation evaluation benchmark that covers the core capabilities of robot models across multiple dimensions and in a comprehensive manner.

https://github.com/AgibotTech/genie_sim

https://huggingface.co/datasets/agibot-world/GenieSim3.0-Dataset/tree/main/task_suite

https://modelscope.cn/datasets/agibot_world/GenieSim3.0-Dataset/tree/master/task_suite