According to a report from Economic Daily News, citing Chinese media outlet Jiemian News, Xiaomi is working on building its own 10,000 GPU cluster and currently heavily investing in AI large models. The company established its AI large model team in April 2023, with 1,200 R&D personnel led by CEO Lei Jun. At that time, Xiaomi had already demonstrated substantial commitment to AI development, with access to 6,500 GPUs, as noted by the report.
In May 2024, Xiaomi announced that its large language model “MiLM” officially passed the large model filing requirements. Also, Lei Jun previously stated that Xiaomi’s smart voice assistant XiaoAi has 120 million monthly active devices, as cited by the report from Economic Daily News.
In August 2023, Zeng Xuezhong, Xiaomi Group Senior Vice President and President of the Mobile Phone Department, revealed that based on their deep technical cooperation with Qualcomm and MediaTek, Xiaomi would soon launch AI large model applications.
According to Jiemian News, Lei Jun has emphasized a strategic focus on mobile phone development over alternative hardware solutions such as glasses, emphasizing Xiaomi’s complete commitment to this direction.
During his 2023 annual speech, Lei Jun announced that Xiaomi had successfully run demos on mobile phones with on-device large models. The report notes that these models, operating at 1.3 billion parameters, achieved cloud-comparable performance in some scenarios. Additionally, the company also developed a more extensive model reaching 6 billion parameters.
Lei Jun stated that Xiaomi built its first visual AI team in 2016 and has established a dedicated large model team in April 2023. As noted by the report, during six expansions in seven years, Xiaomi’s AI team gradually built AI technological capabilities in vision, speech, acoustics, knowledge graphs, NLP, machine learning, and multimodal systems.
Notably, the report mentions that unlike other companies, Xiaomi’s main direction for large model technology focuses on lightweight implementation and local deployment.
Read more
(Photo credit: Xiaomi)