Spirit AI, a rising Chinese artificial intelligence firm, has announced that its latest foundation model, Spirit v1.5, has secured the first-place ranking on the RoboChallenge global robotics benchmark. The model notably outperformed a leading U.S.-based counterpart, signaling a major breakthrough for China in the field of "embodied intelligence"—AI that operates within physical robotic bodies, according to Xinhua.
Setting New Industry Records According to the latest RoboChallenge leaderboard, Spirit v1.5 earned a total score of 66.09. It achieved a task success rate of 50.33 percent, making it the only model currently evaluated to surpass the 50 percent threshold. This performance placed it ahead of the pi0.5 model, developed by the prominent American company Physical Intelligence.
RoboChallenge is widely regarded in the tech industry as a "global exam" for robots. It subjects AI models to 30 real-world physical tasks, including precise object placement, tool usage, and target recognition, to test their ability to function in everyday environments.
Innovation through Open Source In a move to foster further development, Spirit AI has open-sourced Spirit v1.5 and its associated resources. The model utilizes a Vision-Language-Action (VLA) architecture. Unlike traditional modular systems that separate perception from action, this end-to-end system integrates reasoning and execution into a single workflow, significantly reducing operational errors.
A Hub for Robotics Excellence Spirit AI is headquartered in Hangzhou, a city rapidly becoming a global AI and robotics epicenter, also serving as the home for companies like DeepSeek and Unitree Robotics. Spirit AI previously made headlines in June 2025 with the launch of its Moz1 humanoid robot, which is designed for industrial and logistics applications.
Experts believe that while embodied intelligence is not yet ready for mass-market deployment, Spirit v1.5’s success represents a vital step toward making service robots a practical reality. Spirit AI leadership anticipates a significant surge in the availability of capable service robots within the next two to three years.
Follow the latest news on Google News
