Former DeepSeek researcher Luo Fuli has officially joined XIAOMI-W. On November 12, it was revealed that Luo posted on social media: "Intelligence will eventually evolve from language to the physical world. At Xiaomi MiMo, I’m working alongside a team of creative, talented, and passionate researchers to build this future—fully committed to realizing our vision of AGI."

Xiaomi MiMo is XIAOMI-W's first large-scale inference model. This announcement confirms Luo’s move to the company and addresses prior industry speculation.

Earlier reports suggested that XIAOMI-W founder Lei Jun had recruited Luo—one of the key developers behind DeepSeek’s open-source model DeepSeek-V2—with an annual salary in the tens of millions to lead the company’s AI model team. Concurrently, XIAOMI-W was reportedly building a 10,000-GPU cluster to bolster its AI capabilities. The team initially had access to 6,500 GPUs, and this expansion aims to strengthen its R&D capacity in large-scale AI models. Sources indicate the project has been underway for months, with Lei playing a pivotal leadership role. He views smartphones as central to AI hardware and has prioritized significant investment in this area.

Neither XIAOMI-W nor Luo had previously commented on the rumors. On February 19, Luo posted on social media requesting privacy, stating, "I’m no prodigy—just someone quietly working on difficult but meaningful things." Earlier reports on February 18 claimed she had started at a new company, but internal XIAOMI-W records at the time did not list her name.

Notably, in October, XIAOMI-W’s AI team collaborated with Peking University on a research paper focusing on Mixture of Experts (MoE) and reinforcement learning, which included Luo’s name—suggesting her formal debut in her new role.

Luo holds a bachelor’s degree in computer science from Beijing Normal University and a master’s in computational linguistics from Peking University. After graduating, she joined Alibaba DAMO Academy as a researcher in machine intelligence, contributing to the multilingual pre-trained model VECO and the open-sourcing of the AliceMind project. In 2022, she joined DeepSeek’s parent company, working on deep learning before becoming a researcher at DeepSeek, where she helped develop models like DeepSeek-V2.