Xiaomi has open-sourced a new foundation model that integrates autonomous driving and embodied artificial intelligence, as the Chinese smartphone and electric-vehicle (EV) manufacturer seeks to enhance its AI capabilities.
MiMo-Embodied, which comes after Xiaomi’s first large language model MiMo launched in April, was described as the “first open-source vision-language model” that combined autonomous driving and embodied AI tasks. Technical reports and project files are available on platforms such as Hugging Face and GitHub.
The cross-embodied model achieved state-of-the-art performance in both domains, according to Xiaomi It demonstrated “superior” results in task planning, affordance prediction and spatial understanding for embodied AI, while excelling in environmental perception, status prediction and drive planning for autonomous driving.
Interest in embodied intelligence, which integrates AI into physical systems, has surged alongside rapid advancements in humanoid robotics in China.
Xiaomi debuted prototypes of a quadrupedal robot dog and a humanoid robot in 2021 and 2022, respectively, and has been heavily investing in related areas.
Xiaomi’s announcement on Friday followed the recent hiring of Luo Fuli, a former core researcher at Chinese AI start-up DeepSeek who joined Xiaomi several months after the company reportedly approached her.
