The labs behind the people.
Use this as the institutional index behind the people deck: who each team is, where it sits, what it is known for, and how its money, model, city, and source trail connects back to the field.
Lab profiles
DeepSeek (深度求索)
Hangzhou foundation-model lab spun off from quant fund High-Flyer (幻方量化). R1 made DeepSeek the cost-curve reference point for open reasoning models, while commercialization and revenue remain intentionally opaque.
Zhipu AI / Z.ai (智谱AI)
Spun out of Tsinghua's Knowledge Engineering Group (KEG); founded by professors Tang Jie (唐杰) and Li Juanzi (李娟子); CEO Zhang Peng (张鹏).
Moonshot AI / Kimi (月之暗面)
Founded by ex-Tsinghua KEG / CMU PhD Yang Zhilin (杨植麟). Long-context flagship Kimi consumer chatbot.
MiniMax (稀宇科技)
Shanghai foundation-model lab founded by ex-SenseTime executives.
StepFun (阶跃星辰)
Shanghai foundation-model lab founded by ex-MSRA Chief Scientist Jiang Daxin (姜大昕). Densest MSRA-diaspora cluster. Yin Qi (印奇, Megvii co-founder) became chairman January 2026.
Qwen team @ Alibaba (通义千问 / 通义实验室)
Tongyi Lab established end of 2022 by merging DAMO language/vision/AI teams into Alibaba Cloud. 36Kr and TMTPost frame Qwen as Alibaba's model layer for cloud AI revenue, Bailian MaaS, open-source distribution, and agent/coding benchmarks; Qwen itself has no standalone valuation or P&L.
ByteDance Doubao + Seed (豆包 / Seed)
ByteDance's Doubao/Seed foundation-model org (Beijing + Bay Area). Wu Yonghui (吴永辉, ex-Google DeepMind VP) joined as Seed lead February 2025; recent coverage emphasizes Doubao's consumer feedback loop, Seedance video, Volcano Engine token share, and ByteDance's experiments around phone/on-device agent access.
Hunyuan team @ Tencent (混元)
Tencent's foundation-model org. Yao Shunyu (姚顺雨, 28, ex-OpenAI) appointed Chief AI Scientist December 2025. Tencent AI Lab dissolved into Hunyuan March 2026. Recruited Microsoft's WizardLM team in 2025.
Xiaomi MiMo (小米 MiMo)
Xiaomi's foundation-model team within the AI Lab. Established April 2023; rebranded as 'MiMo' around its 2025 open-source release. Luo Fuli (罗福莉, ex-DeepSeek-V2 core developer) became team head November 2025.
Meituan LongCat (美团龙猫)
Meituan-owned LongCat large-model family. LongCat-Flash-Chat was released and open-sourced September 1, 2025 as a 560B MoE model that activates roughly 18.6B-31.3B parameters per token and emphasizes efficient inference, agent/tool-use, and local-services workflows.
Baichuan AI (百川智能)
Founded by ex-Sogou CEO Wang Xiaochuan (王小川). Pivoted to healthcare AI mid-2024; flagship is 'Futang Baichuan' pediatric AI model with Beijing Children's Hospital.
Where these labs came from
Not foundation-model labs themselves — the quant funds, vision companies, search teams, and university labs whose people and spinouts seeded today's frontier labs.
01.AI (零一万物)
Founded by Kai-Fu Lee (李开复), incubated by Sinovation Ventures. Exited frontier pre-training Q1 2025.
High-Flyer (幻方量化)
Quant hedge fund co-founded by Liang Wenfeng. AI division established 2019; spun off DeepSeek in 2023.
Megvii (Face++) (旷视科技)
Computer vision company founded by Tsinghua Yao Class trio (Yin Qi, Tang Wenbin, Yang Mu). Important predecessor lab — Yin Qi went on to chair StepFun.
Recurrent AI (循环智能)
Yang Zhilin and Zhang Yutao's first venture before Moonshot.
SenseTime (商汤科技)
Computer vision unicorn. SenseTime alumni cluster founded MiniMax (Yan Junjie + Zhou Yucong + Yun Yeyi).
Sogou (搜狗)
Search engine spun off from Sohu by Wang Xiaochuan. Acquired by Tencent 2021. Wang plus 3 ex-Sogou execs founded Baichuan in 2023.
Tsinghua KEG (Knowledge Engineering Group) (清华大学知识工程实验室)
Tsinghua CS lab led by Tang Jie + Li Juanzi. Direct origin of Zhipu (2019) and Yang Zhilin's training (Moonshot).