Post-95 "AI genius" Luo Fuli's debut: The gap between Chinese models and top-tier ones has shrunk from three years to just a few months! She comes from an ordinary rural family and was once poached by Lei Jun with an annual salary of ten million yuan

This morning, the 2025 Xiaomi "Human-Vehicle-Home Full Ecology" Partner Conference was held in Beijing, with reporters from the Daily Economic News attending the event. At the conference, Xiaomi officially announced the open-source launch of the Xiaomi MiMo-V2-Flash large model. Meanwhile, it was noted that Luo Fuli, the head of the MiMo large model, made her first public appearance at this event. This 95-born technical expert, known in the industry as an "AI prodigy," introduced the specifics of Xiaomi's large model and shared her views on AI.

Image source: Daily Economic News reporter Yang Hui

At the event, Luo Fuli appeared somewhat nervous, focusing primarily on technical interpretations. She mentioned that the research and development initially revolved around three questions: the intelligent agent must have an efficient communication language, namely coding ability and tool invocation ability; the current communication bandwidth between intelligent agents is very low, how to accelerate the bandwidth; and how to stimulate post-training potential.

Image source: Xiaomi "Human-Vehicle-Home Full Ecology" Partner Conference

According to Luo Fuli, the Xiaomi MiMo-V2-Flash is a very small model, with a total of only 309 billion parameters and only 15 billion active parameters. "I wouldn't even call it a very large model. However, in terms of its coding ability, I believe it has entered the global top 2."

Luo Fuli stated that the Xiaomi MiMo-V2-Flash surpasses DeepSeek V3.2 and K2-Thinking on most evaluation benchmarks, while the parameter count has been reduced by half to two-thirds. In the global top models that are roughly at the same level, MiMo-V2-Flash achieves low cost and high speed, and has initially demonstrated the ability to simulate the world.

Image source: Daily Economic News reporter Yang Hui

Luo Fuli believes that the next generation of intelligent agent systems is not a "language simulator," but a true "agent" that understands the human world and coexists with it. In terms of execution capabilities, it should achieve a transition from "answering questions" to "completing tasks," possessing abilities such as memory, reasoning, autonomous planning, decision-making, and execution. From the perspective of omni-perception, it should achieve unified multi-modal perception, laying the foundation for understanding the physical world, thereby better integrating into smart terminals like glasses Luo Fuli also recalled her experience when she first entered the large model field in 2020. "At that time, the gap between domestic open-source models and the world's top closed-source models was at least three years. But now everyone can see that China's open-source models, like DeepSeek and MiMo, are only a few months away from the world's top closed-source models," Luo Fuli said.

Speaking about the value of open-source, Luo Fuli stated, "We believe that the value of open-source is not just about sharing models and sharing code. I think open-source is the only way to achieve the popularization of AGI and ensure the co-evolution of all human wisdom."

Born after 1995, Luo Fuli attended the "Qingbei Class" at Yibin No. 1 High School in Sichuan Province. She studied computer science at Beijing Normal University for her undergraduate degree and graduated with a master's degree in computational linguistics from the Institute of Computational Linguistics at Peking University. During her studies, she published eight papers at the top international conference in artificial intelligence, ACL, in 2019, two of which were as the first author.

▲Luo Fuli Photo from Luo Fuli's WeChat Official Account

Luo Fuli carries many labels: she was recommended to Peking University, published articles in top conferences and journals, graduated and joined Alibaba DAMO Academy, switched to a well-known private equity firm, engaged in artificial intelligence research at DeepSeek, and became an overnight sensation on Zhihu... She has thus been referred to by netizens as an "AI prodigy."

Her career began at Alibaba DAMO Academy, where she led the development of the multilingual pre-training model VECO and promoted the open-source work of AliceMind. In 2022, Luo Fuli joined the parent company of DeepSeek, Huanfang Quantitative, to work on deep learning-related tasks, and later became a deep learning researcher at DeepSeek, participating in the development of models like DeepSeek-V2.

According to Hongxing News, a person familiar with Luo Fuli's family situation from the Yibin Luo Clan Research Association told reporters that Luo Fuli was born into an ordinary family in a village in Yibin City, with her father being an electrician and her mother a teacher. Her younger sister is now a doctor. "A few years ago, Luo Fuli's uncle said she was doing foundational design in artificial intelligence, and I couldn't believe that our Yibin Luo Clan had such high-end technology talent, but I found Luo Fuli's name in the Peking University alumni directory and later saw this academic star on the Peking University official website. She is indeed the pride of the Luo family."

It is understood that the village where Luo Fuli was born and raised has a strong reading atmosphere. When she was young, an elderly man named Luo Pengyuan, who retired from a certain agency in Yibin City, opened a rural bookstore in the village, providing books and tutoring for free until he passed away in 2020.

Earlier this year, according to the Securities Times, Xiaomi founder Lei Jun reportedly hoped to poach Luo Fuli with an annual salary of tens of millions, inviting her to Xiaomi to lead a team in AI large model research. This news once trended on Weibo. Regarding Lei Jun's invitation, it was reported that at that time, Luo Fuli was still considering whether to join Xiaomi On November 12th, Luo Fuli posted on her social media: "Intelligence will eventually move from language to the physical world. I am at Xiaomi MiMo, working with a group of creative, talented, and genuinely passionate researchers to build such a future, fully committed to our vision of AGI." Xiaomi MiMo is Xiaomi's first reasoning large model. This also marks Luo Fuli's official announcement of her joining Xiaomi, responding to previous rumors.

In early December, Lu Weibing indirectly responded to Luo Fuli's joining through a video, stating that "Xiaomi craves and respects top talent."

According to the timeline released by Lu Weibing for Xiaomi's self-developed MiMo series, Xiaomi has already launched the reasoning large model MiMo-7B, the visual reasoning large model MiMo-VL, the native end-to-end audio generation model MiMo-Audio, the edge-side visual language large model MiMo-VL-Miloco, and the embodied large model MiMo-Embodied.

Lu Weibing revealed that Xiaomi has initiated "intensive investment" in the AI field, with the progress of large models and applications "far exceeding expectations," and will focus on the core direction of "deep integration of AI and the physical world" in the future. Xiaomi hopes to provide a good development platform for outstanding talents.

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this is at one's own risk