Yan Junjie
Last reviewed
Sources
18 citations
Review status
Source-backed
Revision
v3 · 1,805 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
18 citations
Review status
Source-backed
Revision
v3 · 1,805 words
Add missing citations, update stale details, or suggest a clearer explanation.
Yan Junjie (闫俊杰, born 1989) is a Chinese artificial intelligence researcher and entrepreneur who is the founder, chairman, chief executive officer, and chief technology officer of MiniMax, the Shanghai-based AI company behind the abab and MiniMax-M foundation models and the Hailuo video and audio products.[1][2] A former vice president of SenseTime, he founded MiniMax in December 2021 and took it public on the Hong Kong Stock Exchange on January 9, 2026, a debut that more than doubled the offer price and made him, at 36, one of China's youngest AI billionaires.[2][6]
Yan Junjie is the founder, chairman, CEO, and CTO of MiniMax, the Shanghai-based AI company known for its abab and MiniMax-M foundation models, the Hailuo video generator, and the AI companion apps Talkie and Xingye.[1][2] A former vice president of SenseTime, where he helped lead deep learning research for more than six years, Yan founded MiniMax in December 2021 around the conviction that general-purpose AI should be built for the broad public.[1][3] He took the company public on the Hong Kong Stock Exchange on January 9, 2026, in a debut that more than doubled the offer price and made him a billionaire.[2][6] Ten days later he became the second founder of a Chinese foundation-model company, after DeepSeek's Liang Wenfeng, to address a symposium chaired by Premier Li Qiang.[13][14]
Yan was born in 1989 in Henan province. Accounts of his school years describe him as a strong mathematics student who taught himself advanced calculus in high school because the standard curriculum moved too slowly.[2] He received a bachelor's degree in mathematics from Southeast University in Nanjing in June 2010 and a doctorate in artificial intelligence from the Institute of Automation of the Chinese Academy of Sciences in July 2015, after which he conducted postdoctoral research in the computer science department of Tsinghua University, focusing on deep learning and computer vision.[1][2]
Yan joined SenseTime in 2015, reportedly starting as an intern, and spent more than six years at the computer vision company, rising to vice president and vice-head of its research institute.[1][2] He oversaw work spanning deep learning technology, distributed computing, inference systems, and general intelligence research, and Zhou Yucong, a later MiniMax colleague, led algorithm R&D under the same umbrella.[3] According to MiniMax's listing documents, Yan has published roughly 200 academic papers that have drawn more than 30,000 citations, and his research honors include two Wu Wenjun Artificial Intelligence awards in October 2019 and a first prize in the Guangdong Province Technology Invention Award in February 2020.[1]
An avid gamer, Yan took notice of OpenAI in 2019 after its agents defeated elite human teams in the strategy game Dota 2, an episode that pushed his interests from computer vision toward language models and general-purpose AI.[2][8] He left SenseTime at the end of 2021, just as the company was preparing its own Hong Kong listing, later explaining: "We believed that building general-purpose AI for the public was important. The aim was to develop a model company that could also make good products."[3]
Yan founded MiniMax, formally MiniMax Group, in Shanghai in December 2021, naming it after the minimax algorithm from game theory.[3] The early team drew heavily on former SenseTime researchers, including Zhou Yucong, who became the company's visual model research lead; some accounts also name Yang Bin among the co-founders.[1][3] The video game developer miHoYo provided early backing, and subsequent investors included Tencent, HongShan, Hillhouse, IDG Capital, and ZhenFund, which participated in six consecutive rounds.[5] In March 2024, Alibaba led a US$600 million financing that valued MiniMax at about US$2.5 billion, the company's second major backing from Alibaba that year.[4]
MiniMax is one of China's so-called "AI tigers," a cohort of well-funded large-model startups built to rival U.S. labs such as OpenAI.[6] Unusually for a Chinese model lab, it paired frontier model research with consumer products aimed at global markets from the start. By the end of 2025 the company reported more than 236 million cumulative users across over 200 countries and regions, around 214,000 enterprise customers and developers, and more than 70 percent of revenue from outside China.[16] In July 2025, NVIDIA CEO Jensen Huang publicly cited MiniMax among world-class Chinese AI innovators.[3]
MiniMax's first consumer product, the AI companion app Glow, launched in October 2022 and reached roughly 5 million users within four months before being removed from Chinese app stores in March 2023 over regulatory requirements.[3] The company relaunched the concept as Talkie for international markets in June 2023 and Xingye for the domestic market in September 2023.[3] Talkie, a rival to Character.AI, reached 11 million monthly active users by July 2024, more than half of them in the United States, before being pulled from the US iOS App Store in December 2024; MiniMax cited technical reasons amid wider scrutiny of Chinese-owned apps.[12]
The company's foundation models evolved from the proprietary abab family into the open-weight MiniMax-01 and M series:
| Release | Model | Notes |
|---|---|---|
| 2023 | abab series | Proprietary LLM family powering MiniMax's apps and open platform [9] |
| January 2024 | abab6 | First abab model on a mixture-of-experts architecture [9] |
| April 2024 | abab 6.5 / 6.5s | Trillion-parameter-scale MoE with 200,000-token context [9] |
| January 2025 | MiniMax-01 (Text-01, VL-01) | Open-weight MoE, 456B total / 45.9B active parameters, lightning attention [10] |
| June 2025 | MiniMax-M1 | Described as the first open-weight large-scale hybrid-attention reasoning model; 1 million-token context; Apache 2.0 license [10] |
| October 2025 | MiniMax-M2 | Compact open-weight MoE focused on coding and agentic workflows [6] |
In early 2026, MiniMax shipped further M-series updates, including MiniMax-M2.5.[18] On the multimodal side, the Hailuo line began with the video-01 text-to-video model in September 2024, whose lifelike human motion drew viral attention, followed by Hailuo 02 in mid-2025 and Hailuo 2.3 in late 2025; the company also markets speech synthesis (Speech-02 and later Speech 2.6, under the MiniMax Audio banner) and Music models.[6][11] In September 2025, Disney, Universal, and Warner Bros. Discovery sued MiniMax in the United States, alleging copyright infringement in connection with Hailuo-generated content; the case was pending at the time of the IPO.[17]
MiniMax filed for a Hong Kong listing in 2025 and priced its offering at HK$165 per share, the top of its range, with 14 cornerstone investors including the Abu Dhabi Investment Authority, Alibaba, Aspex, Eastspring, Mirae Asset, and E Fund; the Hong Kong retail tranche was oversubscribed about 1,837 times.[5] The stock debuted on January 9, 2026 under the code 0100, days after rival Zhipu AI, making MiniMax the second major Chinese large-model developer to go public.[5][6] The deal raised roughly HK$4.8 billion (about US$618 to 620 million), with proceeds of up to about US$710 million assuming full exercise of the over-allotment option.[5][6][17] Shares closed their first day at HK$345, up 109 percent, lifting the company's market value above HK$100 billion (about US$12.8 billion).[6][7]
Yan controlled roughly a quarter of the company (about 24 to 25 percent, held through four holding vehicles), a stake worth around US$1.6 billion at the offer price and approximately US$3.2 billion after the first-day surge, placing him among China's youngest AI billionaires at age 36.[2][8] For full-year 2025, MiniMax reported revenue of US$79.0 million, up 158.9 percent year over year, with an adjusted net loss of US$250.9 million and cash reserves above US$1 billion; under IFRS its net loss for the first nine months of 2025 was US$512 million.[5][16] On January 19, 2026, Yan spoke at a symposium chaired by Premier Li Qiang on the draft Government Work Report and the 15th Five-Year Plan, a slot read as official endorsement of MiniMax's globally oriented model business.[13][14]
Yan's stated mission for MiniMax is "Intelligence with Everyone," a phrase displayed in the company's offices that he describes as "the founding vision that drives everything we do."[15] In a keynote titled "Everyone's AI" at the World Artificial Intelligence Conference in Shanghai on July 30, 2025, he argued that artificial general intelligence is achievable but "must serve the many, not the few," that it will be built together with users rather than by any single company, and that it "should belong to all of us."[15] He pointed to diverse alignment objectives, multi-agent systems, and accelerating open-source development as forces working against monopolization of AI, and noted that inference costs for leading models had fallen by an order of magnitude in a year.[15]
Yan has resisted the "China's OpenAI" label, framing AGI as inherently multimodal and treating AI progress as an engineering problem tractable through first-principles methodology rather than the work of a few geniuses.[2][3] Following MiniMax's 2025 results, he said the firm would "evolve from a large-model company into a platform company for the AI era."[16]