In June 2026, Spirit AI, a physical AI start-up based in Hangzhou, Zhejiang province and founded in 2024, briefly overtook US tech giant Nvidia to claim the top spot on RoboArena, a global benchmark for physical AI, with its Spirit v1.6 model launched earlier that month.

RoboArena, co-developed by Nvidia alongside Stanford University and the University of California, Berkeley, assesses how effectively generalist robot policies translate digital commands into real-world robotic actions. The benchmark is often described as the “Olympics” of embodied intelligence in North America.

However, just days after Spirit AI’s rise, the RoboArena team overhauled its evaluation methodology and removed Spirit’s model from the official rankings. Several other competitors, including a model from Chinese start-up X Square Robot which had ranked fourth, were also dropped following findings of "benchmark hacking."

Pranav Atreya, a lead author of the RoboArena project and PhD student at UC Berkeley, stated in a June post on X that the team had “retroactively removed evaluations from organisations who [it] found to be engaging in benchmark manipulation.”

This controversy highlights the intense competition between the US and China in developing next-generation artificial intelligence and the challenges in fairly evaluating autonomous systems.

Sources