Xiaomi (1810) unveiled its self-developed artificial intelligence model "Xiaomi MiMo-V2-Flash" on Wednesday, with "AI prodigy" Luo Fuli making her first public introduction for the company.
In addition to supporting deep thinking and online search functions, this model could chat and is useful in scenarios where real-time information, the latest updates, or data verification are needed.
Luo said that from her perspective, Xiaomi's open-source model's code and agent capabilities have already entered the top 1 or 2 globally in very open and fair evaluation lists.
She noted that the newly launched AI large model highlights strong capabilities in code and tool invocation, incorporates a model structure designed for extreme inference efficiency, and introduces a new post-training paradigm for efficient and stable extended reinforcement learning training.
Besides, this model boasts extreme inference efficiency. In the top-tier model speed and cost quadrant globally, it achieves low cost and high speed.
Luo, who is in her 20s, previously worked as a deep learning researcher at DeepSeek, and was one of the core developers of the domestically developed large-scale model DeepSeek-V2. She announced that she has joined Xiaomi in November.
Lu Weibing, partner and president of Xiaomi, mentioned that Xiaomi will invest 200 billion yuan (HK$220.9 billion) in research and development over the next five years, with the long-term goal of becoming a global leader in core technologies.
