深翼情报台一号 · AI 架构情报
研究实验室登记册
53 家 · 7 个待补字段 · 登记册核查 2026-10-2253 labs · 7 fields to fill · register checked 2026-10-22
在前沿 AI 研究系统上处于全球前沿的研究实验室登记册。每家一行:所属机构、所在地、研究方向与关键人物,点开一行看官方页与交叉链接。缺的字段直接标「未列」,不省略。
AI 系统与架构AI Systems & Architecture登记册Register跨区:AI 基础设施与经济Cross-tag: AI Infrastructure & Economics阅读时间:约 13 分钟Reading time: About 12 min
- 登记范围Scope
- 在前沿 AI 研究系统上处于全球前沿的研究实验室。每家一行:所属机构、所在地、研究方向与关键人物,附官方页与站内交叉链接。Research laboratories at the global frontier of AI research systems, one row each: institution, location, research focus and key people, with the official site and on site cross links.
- 登记方法Method
- 逐家登记并核对官方页;每行附官方页与交叉链接(交叉链接指向前沿公司名单);字段缺失直接标「未列」,不省略、不推测填补。Each laboratory is registered and checked against its official page, and each row carries the official site and a cross link to the Frontier Companies roster. Missing fields are marked "Not listed", never omitted or filled by inference.
- 证据等级Evidence grade
- 多源已核Multi-source verified
- 核查截止日期Verified through
- 2026-10-222026-10-22
- 登记结论Register findings
- ① 在册 53 家:企业实验室 26 家、高校与研究机构 20 家、独立机构 7 家;地区分布为美国 22 家、中国 21 家、欧洲 4 家、加拿大 4 家、日本 1 家、澳大利亚 1 家。② 收录口径不止模型研发机构:Epoch AI、METR、Artificial Analysis 三家测量与评估机构在册,因为本台引用能力与成本数据时,须回到它们的测量口径对账。③ 全册现有 7 个待补字段,全部明标「未列」。1. The register holds 53 laboratories: 26 corporate labs, 20 universities and research institutes, and 7 independent bodies. By region: 22 in the US, 21 in China, 4 in Europe, 4 in Canada, 1 in Japan and 1 in Australia. 2. The scope is wider than model builders: Epoch AI, METR and Artificial Analysis, three measurement and evaluation bodies, are included because this desk's capability and cost figures must be reconciled against their measurement calibers. 3. Seven fields across the register are still pending, every one of them marked "Not listed".
- 登记局限Limitations
- 覆盖以公开可查的前沿研究实验室为限,册外实验室不在本册范围;关键人物与研究方向随时间变化,以各家官方页为准;待补字段补齐前,相关行只能按「未列」引用。Coverage is limited to publicly documented frontier research laboratories; laboratories outside the register are outside its scope. Key people and research focuses change over time; each laboratory's official page governs, and pending fields can only be cited as "Not listed" until they are completed.
- 后续观察Watch next
- 7 个待补字段逐个补齐;各家关键人物与研究方向的变动随观察节奏更新;达到前沿标准的新实验室按登记口径增补入册。Completion of the seven pending fields; updates to key people and research focuses on the standing watch cadence; new laboratories added under the same registration caliber as they reach the frontier.
| 实验室 · 所属机构Lab · Institution | 所在地Location | 研究方向Focus | 关键人物Key people | |
|---|---|---|---|---|
| 谷歌深度思维Google DeepMind同时在前沿公司名单Also on the Frontier Companies rosterAlphabet(Google)Alphabet (Google)企业实验室Corporate | 英国伦敦与美国山景城London, UK and Mountain View, US | 前沿模型Frontier models科学智能AI for science | Demis Hassabis、Koray Kavukcuoglu、Shane LeggDemis Hassabis, Koray Kavukcuoglu, Shane Legg | |
| 研究方向Focus:通用大模型、强化学习与科学智能(AlphaFold 系)的全栈前沿研究Full-stack frontier research on general large models, reinforcement learning, and science intelligence (the AlphaFold line)官方页Official site:deepmind.google交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| OpenAI 研究部OpenAI Research同时在前沿公司名单Also on the Frontier Companies rosterOpenAI企业实验室Corporate | 美国旧金山San Francisco, US | 前沿模型Frontier models安全与对齐Safety & alignment | Jakub Pachocki、Mark Chen、Sam AltmanJakub Pachocki, Mark Chen, Sam Altman | |
| 研究方向Focus:大模型预训练、推理模型与对齐安全的核心研究组织The core research organization for large-model pretraining, reasoning models, and alignment safety官方页Official site:openai.com交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| Anthropic 研究部Anthropic Research同时在前沿公司名单Also on the Frontier Companies rosterAnthropic企业实验室Corporate | 美国旧金山San Francisco, US | 安全与对齐Safety & alignment智能体Agents | Jared Kaplan、Chris Olah、Dario AmodeiJared Kaplan, Chris Olah, Dario Amodei | |
| 研究方向Focus:可解释性、对齐科学与长程智能体能力的前沿研究Frontier research on interpretability, alignment science, and long-horizon agent capability官方页Official site:anthropic.com交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| Meta 基础人工智能研究院Meta FAIR (Fundamental AI Research)同时在前沿公司名单Also on the Frontier Companies rosterMeta(MSL 体系)Meta (within the MSL system)企业实验室Corporate | 美国门洛帕克与纽约、巴黎等Menlo Park, New York, Paris and more | 基础研究Basic research世界模型World models | 未列Not listed | |
| 研究方向Focus:Meta 的基础研究旗舰,覆盖世界模型、语音视觉与开源基础模型研究Meta's basic-research flagship, covering world models, speech and vision, and open foundation-model research官方页Official site:ai.meta.com交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| Meta 超级智能实验室Meta Superintelligence Labs (MSL)同时在前沿公司名单Also on the Frontier Companies rosterMeta企业实验室Corporate | 美国门洛帕克Menlo Park, US | 前沿模型Frontier models | Alexandr Wang、Shengjia Zhao 赵晟佳Alexandr Wang, Shengjia Zhao (赵晟佳) | |
| 研究方向Focus:2025 年组建的前沿模型总部,统筹下一代旗舰模型与应用研究The frontier-model headquarters formed in 2025, coordinating next-generation flagship models and applied research官方页Official site:ai.meta.com交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 微软研究院Microsoft ResearchMicrosoft企业实验室Corporate | 美国雷德蒙德与全球多地Redmond, US and sites worldwide | 基础研究Basic research | 未列Not listed | |
| 研究方向Focus:大厂综合研究院,覆盖机器学习理论、系统与人机交互的长期研究A broad corporate research institute with long-term work across machine-learning theory, systems, and human-computer interaction官方页Official site:microsoft.com/en-us/research | ||||
| 苹果机器学习研究Apple Machine Learning ResearchApple企业实验室Corporate | 美国库比蒂诺与全球多地Cupertino, US and sites worldwide | 多模态Multimodal | 未列Not listed | |
| 研究方向Focus:苹果的公开发表型研究组织,主攻基础模型、多模态与端侧智能Apple's publishing research organization, focused on foundation models, multimodality, and on-device intelligence官方页Official site:machinelearning.apple.com | ||||
| IBM 研究院IBM ResearchIBM企业实验室Corporate | 美国约克敦高地与全球多地Yorktown Heights, US and sites worldwide | 智能体Agents | 未列Not listed | |
| 研究方向Focus:老牌企业研究院,现聚焦基础模型、企业智能体与量子计算A long-established corporate institute, now focused on foundation models, enterprise agents, and quantum computing官方页Official site:research.ibm.com | ||||
| 英伟达研究院NVIDIA ResearchNVIDIA企业实验室Corporate | 美国圣克拉拉与全球多地Santa Clara, US and sites worldwide | 具身智能Embodied AI系统与算力Systems & compute | Tianqi Chen 陈天奇、黄仁勋 Jensen HuangTianqi Chen (陈天奇), Jensen Huang (黄仁勋) | |
| 研究方向Focus:图形与加速计算出身的研究院,现覆盖生成式模型、具身智能与系统架构An institute rooted in graphics and accelerated computing, now covering generative models, embodied intelligence, and systems architecture官方页Official site:research.nvidia.com | ||||
| xAI同时在前沿公司名单Also on the Frontier Companies rosterxAI(2026 年 2 月并入 SpaceX 体系)xAI (folded into the SpaceX system in February 2026)企业实验室Corporate | 美国帕洛阿尔托与孟菲斯Palo Alto and Memphis, US | 前沿模型Frontier models系统与算力Systems & compute | Elon Musk | |
| 研究方向Focus:Grok 系列模型的研究与训练组织,倚重超大规模算力集群The research and training organization behind the Grok model line, built on very large compute clusters官方页Official site:x.ai交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| Mistral AI 研究部Mistral AI Research同时在前沿公司名单Also on the Frontier Companies rosterMistral AI企业实验室Corporate | 法国巴黎Paris, France | 前沿模型Frontier models开源生态Open ecosystem | Arthur Mensch、Guillaume Lample、Timothée LacroixArthur Mensch, Guillaume Lample, Timothée Lacroix | |
| 研究方向Focus:欧洲前沿实验室,以高效开源权重模型与多模态研究立身A European frontier lab built on efficient open-weight models and multimodal research官方页Official site:mistral.ai交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| Sakana AI同时在前沿公司名单Also on the Frontier Companies rosterSakana AI企业实验室Corporate | 日本东京Tokyo, Japan | 科学智能AI for science | Llion Jones、David HaLlion Jones, David Ha | |
| 研究方向Focus:以模型合并、自进化系统与自动化科研为特色的独立前沿实验室An independent frontier lab known for model merging, self-evolving systems, and automated research官方页Official site:sakana.ai交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| Discovery Loop同时在前沿公司名单Also on the Frontier Companies rosterDiscovery Loop(公益公司)Discovery Loop (public benefit corporation)企业实验室Corporate | 美国硅谷Silicon Valley, US | 科学智能AI for science | Jeff Dean、Quoc LeJeff Dean, Quoc Le | |
| 研究方向Focus:2026 年新设实验室,目标是把完整科学实验循环自动化A laboratory founded in 2026, aiming to automate the full scientific experiment loop官方页Official site:未列交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| Thinking Machines Lab同时在前沿公司名单Also on the Frontier Companies rosterThinking Machines Lab企业实验室Corporate | 美国旧金山San Francisco, US | 前沿模型Frontier models | Mira Murati、John Schulman、Lilian Weng 翁荔Mira Murati, John Schulman, Lilian Weng (翁荔) | |
| 研究方向Focus:以后训练与模型定制为核心的新锐前沿实验室A new frontier lab centered on post-training and model customization官方页Official site:thinkingmachines.ai交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 安全超级智能公司Safe Superintelligence (SSI)同时在前沿公司名单Also on the Frontier Companies rosterSafe Superintelligence Inc.企业实验室Corporate | 美国帕洛阿尔托与以色列特拉维夫Palo Alto, US and Tel Aviv, Israel | 安全与对齐Safety & alignment | Ilya Sutskever | |
| 研究方向Focus:单一目标实验室,直指安全超级智能,不做中间产品A single-goal laboratory aiming directly at safe superintelligence, with no intermediate products官方页Official site:ssi.inc交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 深度求索DeepSeek同时在前沿公司名单Also on the Frontier Companies roster深度求索(幻方量化系)DeepSeek (High-Flyer quant family)企业实验室Corporate | 中国杭州Hangzhou, China | 前沿模型Frontier models开源生态Open ecosystem | 梁文锋Wenfeng Liang (梁文锋) | |
| 研究方向Focus:以架构与训练效率创新见长的开源前沿模型实验室An open frontier-model laboratory known for architecture and training-efficiency innovation官方页Official site:deepseek.com交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 月之暗面Moonshot AI同时在前沿公司名单Also on the Frontier Companies roster月之暗面科技有限公司Moonshot AI Technology Co., Ltd.企业实验室Corporate | 中国北京Beijing, China | 前沿模型Frontier models | 杨植麟、周昕宇Zhilin Yang (杨植麟), Xinyu Zhou (周昕宇) | |
| 研究方向Focus:长上下文与注意力架构创新的前沿实验室,Kimi 系列出自此间A frontier lab for long-context and attention-architecture innovation; the Kimi line comes from here官方页Official site:moonshot.ai交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 智谱 Z.AIZ.AI (Zhipu)同时在前沿公司名单Also on the Frontier Companies roster智谱华章(清华 KEG 成果转化)Zhipu Huazhang (spun out of Tsinghua's KEG lab)企业实验室Corporate | 中国北京Beijing, China | 前沿模型Frontier models开源生态Open ecosystem | 唐杰Jie Tang (唐杰) | |
| 研究方向Focus:GLM 系列模型的研发组织,产学研一体的国产开源模型线The R and D organization behind the GLM model line, an academia-industry open model program官方页Official site:z.ai交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 阿里巴巴通义实验室Alibaba Tongyi Lab同时在前沿公司名单Also on the Frontier Companies roster未列Not listed企业实验室Corporate | 中国杭州Hangzhou, China | 前沿模型Frontier models | 未列Not listed | |
| 研究方向Focus:通义千问模型的研究源头,脱胎于达摩院体系The research origin of the Tongyi Qianwen models, grown out of the DAMO Academy system官方页Official site:未列交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 字节跳动 SeedByteDance Seed同时在前沿公司名单Also on the Frontier Companies roster字节跳动ByteDance企业实验室Corporate | 中国北京与多地Beijing and multiple sites, China | 前沿模型Frontier models | 吴永辉Yonghui Wu (吴永辉) | |
| 研究方向Focus:字节的大模型研究体系,支撑豆包全线与火山引擎模型服务ByteDance's large-model research system, behind the Doubao line and Volcano Engine model services官方页Official site:seed.bytedance.com交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| MiniMax 稀宇科技MiniMax同时在前沿公司名单Also on the Frontier Companies roster上海稀宇科技Shanghai MiniMax Technology企业实验室Corporate | 中国上海Shanghai, China | 多模态Multimodal | 闫俊杰Junjie Yan (闫俊杰) | |
| 研究方向Focus:全模态自研的模型实验室,文本、视频、语音与音乐统一栈A model laboratory with a fully self-developed multimodal stack across text, video, speech, and music官方页Official site:minimax.io交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 阶跃星辰StepFun同时在前沿公司名单Also on the Frontier Companies roster上海阶跃星辰智能科技Shanghai StepFun Intelligent Technology企业实验室Corporate | 中国上海Shanghai, China | 多模态Multimodal系统与算力Systems & compute | 朱亦博Yibo Zhu (朱亦博) | |
| 研究方向Focus:以原生多模态统一架构与系统工程为旗帜的前沿实验室A frontier lab built around natively unified multimodal architecture and systems engineering官方页Official site:stepfun.com交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 腾讯混元(基础模型部)Tencent Hunyuan同时在前沿公司名单Also on the Frontier Companies roster腾讯(技术工程事业群)Tencent (Technology Engineering Group)企业实验室Corporate | 中国深圳与北京Shenzhen and Beijing, China | 前沿模型Frontier models | 姚顺雨Shunyu Yao (姚顺雨) | |
| 研究方向Focus:腾讯的大模型研究中枢,混元系列与元宝应用出自此间Tencent's large-model research hub; the Hunyuan line and the Yuanbao app come from here官方页Official site:hunyuan.tencent.com交叉链接Cross-link:前沿公司名单Frontier Companies roster | ||||
| 百度研究院Baidu Research未列Not listed企业实验室Corporate | 中国北京与美国硅谷Beijing, China and Silicon Valley, US | 基础研究Basic research | 未列Not listed | |
| 研究方向Focus:百度早年的海外与基础研究品牌,文心模型与飞桨生态的研究源头之一Baidu's early overseas and basic-research brand, one of the research origins of the Ernie models and the PaddlePaddle ecosystem官方页Official site:未列 | ||||
| 华为诺亚方舟实验室Huawei Noah's Ark Lab未列Not listed企业实验室Corporate | 中国香港与深圳等Hong Kong and Shenzhen, China | 多模态Multimodal | 未列Not listed | |
| 研究方向Focus:华为的中央 AI 研究实验室,2012 年设立,覆盖视觉、语音语义与决策智能Huawei's central AI research laboratory, founded in 2012, covering vision, speech and semantics, and decision intelligence官方页Official site:未列 | ||||
| 商汤科技研究院SenseTime Research商汤科技SenseTime企业实验室Corporate | 中国上海与香港Shanghai and Hong Kong, China | 多模态Multimodal世界模型World models | 未列Not listed | |
| 研究方向Focus:中国视觉智能的发源地之一,现以日日新大模型与世界模型为研究主线One of the birthplaces of computer vision in China, now centered on the SenseNova large models and world models官方页Official site:sensetime.com | ||||
| 斯坦福以人为本人工智能研究院Stanford HAI斯坦福大学Stanford University高校与研究机构University & institute | 美国斯坦福Stanford, US | 基础研究Basic research | Fei-Fei Li 李飞飞Fei-Fei Li (李飞飞) | |
| 研究方向Focus:跨学科 AI 研究院,兼管基础模型研究中心 CRFM 的生态An interdisciplinary AI institute that also stewards the CRFM foundation-model research ecosystem官方页Official site:hai.stanford.edu | ||||
| 麻省理工学院计算机科学与人工智能实验室MIT CSAIL麻省理工学院Massachusetts Institute of Technology高校与研究机构University & institute | 美国剑桥(马萨诸塞)Cambridge, Massachusetts, US | 基础研究Basic research机器人Robotics | Kaiming He 何恺明Kaiming He (何恺明) | |
| 研究方向Focus:全球最大的高校 AI 实验室之一,覆盖学习、机器人与系统One of the world's largest university AI laboratories, spanning learning, robotics, and systems官方页Official site:csail.mit.edu | ||||
| 伯克利人工智能研究实验室Berkeley AI Research (BAIR)加州大学伯克利分校UC Berkeley高校与研究机构University & institute | 美国伯克利Berkeley, US | 机器人Robotics基础研究Basic research | Michael I. Jordan、Stuart Russell、Pieter Abbeel、Sergey LevineMichael I. Jordan, Stuart Russell, Pieter Abbeel, Sergey Levine | |
| 研究方向Focus:开放学术型大实验室,强化学习、机器人与大模型系统重镇A large open academic laboratory, a stronghold for reinforcement learning, robotics, and large-model systems官方页Official site:bair.berkeley.edu | ||||
| 卡内基梅隆大学机器学习系与机器人研究所CMU Machine Learning Department & Robotics Institute卡内基梅隆大学Carnegie Mellon University高校与研究机构University & institute | 美国匹兹堡Pittsburgh, US | 机器人Robotics基础研究Basic research | Jun-Yan Zhu 朱俊彦Jun-Yan Zhu (朱俊彦) | |
| 研究方向Focus:机器学习建系发源地,机器人研究与之并列为双支柱The birthplace of the first machine learning department, with robotics research as its twin pillar官方页Official site:ml.cmu.edu | ||||
| Mila 魁北克人工智能研究所Mila, Quebec AI Institute蒙特利尔大学等魁北克高校共建(非营利)Co-founded by Université de Montréal and other Quebec universities (nonprofit)高校与研究机构University & institute | 加拿大蒙特利尔Montreal, Canada | 基础研究Basic research | Yoshua Bengio | |
| 研究方向Focus:深度学习学术重镇,加拿大国家 AI 战略三所之一A deep-learning academic stronghold, one of the three institutes of Canada's national AI strategy官方页Official site:mila.quebec | ||||
| Vector 研究所Vector Institute多伦多大学等安省高校共建(非营利)Co-founded by the University of Toronto and other Ontario universities (nonprofit)高校与研究机构University & institute | 加拿大多伦多Toronto, Canada | 基础研究Basic research | Geoffrey Hinton | |
| 研究方向Focus:以基础研究与产业落地并重的国家级 AI 研究所A national AI institute balancing basic research with industry application官方页Official site:vectorinstitute.ai | ||||
| 阿尔伯塔机器智能研究所Amii (Alberta Machine Intelligence Institute)阿尔伯塔大学依托(非营利)Anchored at the University of Alberta (nonprofit)高校与研究机构University & institute | 加拿大埃德蒙顿Edmonton, Canada | 基础研究Basic research | Richard Sutton | |
| 研究方向Focus:强化学习的世界中心,加拿大国家 AI 战略三所之一A world center for reinforcement learning, one of the three institutes of Canada's national AI strategy官方页Official site:amii.ca | ||||
| 纽约大学数据科学中心与柯朗研究所NYU Center for Data Science & Courant Institute纽约大学New York University高校与研究机构University & institute | 美国纽约New York, US | 基础研究Basic research | Yann LeCun、Saining Xie 谢赛宁Yann LeCun, Saining Xie (谢赛宁) | |
| 研究方向Focus:LeCun 的学术大本营,覆盖深度学习理论与自监督研究LeCun's academic home base, covering deep-learning theory and self-supervised research官方页Official site:cds.nyu.edu | ||||
| 普林斯顿语言与智能计划Princeton Language and Intelligence (PLI)普林斯顿大学Princeton University高校与研究机构University & institute | 美国普林斯顿Princeton, US | 基础研究Basic research | 未列Not listed | |
| 研究方向Focus:2023 年设立的跨学科计划,聚焦大语言模型的基础与影响An interdisciplinary initiative founded in 2023, focused on the foundations and impact of large language models官方页Official site:pli.princeton.edu | ||||
| 马克斯普朗克智能系统研究所Max Planck Institute for Intelligent Systems马克斯普朗克学会Max Planck Society高校与研究机构University & institute | 德国图宾根与斯图加特Tübingen and Stuttgart, Germany | 基础研究Basic research | 未列Not listed | |
| 研究方向Focus:欧洲机器学习基础研究重镇,ELLIS 网络的发源地之一A European stronghold for basic machine-learning research, and one of the birthplaces of the ELLIS network官方页Official site:is.mpg.de | ||||
| 清华大学交叉信息研究院Tsinghua Institute for Interdisciplinary Information Sciences (IIIS)清华大学Tsinghua University高校与研究机构University & institute | 中国北京Beijing, China | 基础研究Basic research | 姚期智Qizhi Yao (姚期智) | |
| 研究方向Focus:姚期智创立的交叉信息研究院,兼领清华人工智能学院The interdisciplinary information sciences institute founded by Andrew Yao, also leading Tsinghua's College of AI官方页Official site:iiis.tsinghua.edu.cn | ||||
| 清华大学智能产业研究院Tsinghua Institute for AI Industry Research (AIR)清华大学Tsinghua University高校与研究机构University & institute | 中国北京Beijing, China | 机器人Robotics | 张亚勤Ya-Qin Zhang (张亚勤) | |
| 研究方向Focus:面向产业落地的 AI 研究院,聚焦智慧交通、机器人与智能物联An industry-facing AI institute focused on smart transportation, robotics, and intelligent IoT官方页Official site:air.tsinghua.edu.cn | ||||
| 上海人工智能实验室Shanghai AI Laboratory新型研发机构(上海市与高校科研机构共建)A new-type R and D institution, co-built by Shanghai municipality, universities, and research institutes高校与研究机构University & institute | 中国上海Shanghai, China | 开源生态Open ecosystem前沿模型Frontier models | 未列Not listed | |
| 研究方向Focus:国家级新型研发机构,书生 InternLM 与 InternVL 开源模型线的出品方A national-level new-type R and D institution, producer of the InternLM and InternVL open model lines官方页Official site:shlab.org.cn | ||||
| 北京智源人工智能研究院Beijing Academy of Artificial Intelligence (BAAI)北京市支持的新型研发机构A Beijing-supported new-type R and D institution高校与研究机构University & institute | 中国北京Beijing, China | 开源生态Open ecosystem基础研究Basic research | 黄铁军Tiejun Huang (黄铁军) | |
| 研究方向Focus:悟道大模型与智源社区的运营方,北京 AI 学术生态的枢纽Operator of the WuDao model line and the BAAI community, a hub of Beijing's AI academic ecosystem官方页Official site:baai.ac.cn | ||||
| 北京通用人工智能研究院Beijing Institute for General Artificial Intelligence (BIGAI)北京大学与北京市等共建(新型研发机构)Co-founded by Peking University and Beijing municipality (new-type R and D institution)高校与研究机构University & institute | 中国北京Beijing, China | 具身智能Embodied AI | 朱松纯Song-Chun Zhu (朱松纯) | |
| 研究方向Focus:朱松纯创立的通用人工智能专研机构,主攻认知与具身通用智能A dedicated general-AI institute founded by Song-Chun Zhu, focused on cognitive and embodied general intelligence官方页Official site:bigai.ai | ||||
| 中国科学院自动化研究所Institute of Automation, Chinese Academy of Sciences (CASIA)中国科学院Chinese Academy of Sciences高校与研究机构University & institute | 中国北京Beijing, China | 多模态Multimodal基础研究Basic research | 未列Not listed | |
| 研究方向Focus:中国自动化与模式识别的国家队,紫东太初大模型的出品方China's national team for automation and pattern recognition, producer of the Zidong Taichu large models官方页Official site:ia.cas.cn | ||||
| 南京大学 LAMDA 实验室LAMDA, Nanjing University南京大学计算机学院School of Computer Science, Nanjing University高校与研究机构University & institute | 中国南京Nanjing, China | 基础研究Basic research | 周志华Zhi-Hua Zhou (周志华) | |
| 研究方向Focus:周志华领衔的机器学习实验室,中国 ML 学术的标志性团队The machine-learning laboratory led by Zhi-Hua Zhou, a landmark team in Chinese ML academia官方页Official site:lamda.nju.edu.cn | ||||
| 北京大学人工智能研究院Peking University Institute for Artificial Intelligence北京大学Peking University高校与研究机构University & institute | 中国北京Beijing, China | 基础研究Basic research | 林宙辰、朱松纯Zhouchen Lin (林宙辰), Song-Chun Zhu (朱松纯) | |
| 研究方向Focus:北大的校级 AI 研究平台,与智能学院并行运转Peking University's university-level AI research platform, running alongside its school of intelligence官方页Official site:ai.pku.edu.cn | ||||
| 香港科技大学人工智能研究生态HKUST AI Research (incl. HKUST-KnowComp)香港科技大学Hong Kong University of Science and Technology高校与研究机构University & institute | 中国香港Hong Kong, China | 基础研究Basic research | 杨强 Qiang YangQiang Yang (杨强) | |
| 研究方向Focus:港科大的 AI 研究集群,联邦学习与知识计算的传统强项HKUST's AI research cluster, traditionally strong in federated learning and knowledge computing官方页Official site:hkust.edu.hk | ||||
| 香港大学数据科学研究院HKU Institute of Data Science香港大学The University of Hong Kong高校与研究机构University & institute | 中国香港Hong Kong, China | 基础研究Basic research | 马毅 Yi MaYi Ma (马毅) | |
| 研究方向Focus:港大跨学科数据科学与 AI 研究平台HKU's interdisciplinary data-science and AI research platform官方页Official site:ids.hku.hk | ||||
| 艾伦人工智能研究所Allen Institute for AI (AI2)非营利(Paul Allen 创立)Nonprofit (founded by Paul Allen)独立机构Independent | 美国西雅图Seattle, US | 开源生态Open ecosystem科学智能AI for science | 未列Not listed | |
| 研究方向Focus:开放科学路线的非营利实验室,OLMo 开源模型与 AI2-THOR 出自此间A nonprofit laboratory on the open-science path; the OLMo open models and AI2-THOR come from here官方页Official site:allenai.org | ||||
| EleutherAI非营利(社区型研究集体)Nonprofit (community research collective)独立机构Independent | 美国(分布式)United States (distributed) | 开源生态Open ecosystem测量与评估Measurement & evaluation | 未列Not listed | |
| 研究方向Focus:开源大模型与评测的社区型研究集体,lm-eval-harness 的维护方A community research collective for open large models and evaluation, maintainer of lm-eval-harness官方页Official site:eleuther.ai | ||||
| Kyutai非营利(私人捐赠设立)Nonprofit (established by private donation)独立机构Independent | 法国巴黎(及德国图宾根)Paris, France (and Tübingen, Germany) | 开源生态Open ecosystem多模态Multimodal | 未列Not listed | |
| 研究方向Focus:欧洲独立开源实验室,Moshi 实时语音模型为代表作An independent European open laboratory, known for the Moshi real-time speech model官方页Official site:kyutai.org | ||||
| LawZero非营利(慈善捐赠设立)Nonprofit (established by philanthropic donation)独立机构Independent | 加拿大蒙特利尔Montreal, Canada | 安全与对齐Safety & alignment | Yoshua Bengio | |
| 研究方向Focus:Bengio 2025 年创立的安全优先实验室,主推非智能体 Scientist AI 路线A safety-first laboratory founded by Bengio in 2025, advancing the non-agentic Scientist AI path官方页Official site:lawzero.org | ||||
| Epoch AI非营利研究机构Nonprofit research institute独立机构Independent | 美国(旧金山湾区)United States (San Francisco Bay Area) | 测量与评估Measurement & evaluation | 未列Not listed | |
| 研究方向Focus:AI 能力与算力趋势的独立测量机构An independent measurement institute for AI capability and compute trends官方页Official site:epoch.ai | ||||
| METR非营利研究机构Nonprofit research institute独立机构Independent | 美国旧金山San Francisco, US | 测量与评估Measurement & evaluation | 未列Not listed | |
| 研究方向Focus:AI 能力与自主性评估机构,任务时长曲线出自此间An evaluation institute for AI capability and autonomy; the task-length time-horizon curves come from here官方页Official site:metr.org | ||||
| Artificial Analysis独立测量机构Independent measurement institute独立机构Independent | 澳大利亚悉尼Sydney, Australia | 测量与评估Measurement & evaluation | 未列Not listed | |
| 研究方向Focus:模型价格、速度与能力的独立测量方An independent measurer of model prices, speed, and capability官方页Official site:artificialanalysis.ai | ||||
没有匹配的实验室。
Epoch AI、METR、Artificial Analysis 是测量与评估机构,不做前沿模型研发。收录它们,是因为本台引用能力与成本数据时,要回到它们的测量口径对账。

