Workflow
腾讯混元新模型HY 3.0
icon
Search documents
AI Lab撤销,腾讯AI组织架构再变阵
新华网财经· 2026-03-21 05:35
本周早些时候的媒体交流会上,腾讯总裁刘炽平谈到,过去几个月腾讯针对AI密集进行团队调整升级,并进行组织和工作流程重构。随后, 腾讯就传出进行AI组织架构调整的消息。 记者了解,3月20日,腾讯内部通知,撤销AI实验室AI Lab,部分人员调整至大语言模型部,加入混元团队,向首席AI科学家姚顺雨汇报, 部分人员调整至腾讯技术工程事业群(TEG)下属的产学研合作中心。此外,腾讯公司副总裁蒋杰不再兼管AI Lab,其他管理职责不变。 此次AI Lab撤销属于TEG的架构调整。AI Lab是TEG的下一级部门,同样属于TEG下一级部门的还有AI Infra部、大语言模型部。去年12 月,腾讯新成立AI Infra部等部门,以强化大模型研发体系与核心能力,前OpenAI研究员姚顺雨加入腾讯后,出任"CEO/总裁办公室"首席 AI科学家,并兼任AI Infra部、大语言模型部负责人。 TEG是混元大模型团队所在的事业群。此前姚顺雨的加入被视为腾讯加强AI研发能力的信号,他加入后将为腾讯基础大模型混元带来什么改 变也为外界所期待。 混元最近的一个重要更新将在4月。据腾讯介绍,将在4月推出的腾讯混元新模型HY 3.0目前正在 ...
AI Lab撤销,腾讯AI组织架构再变阵
第一财经· 2026-03-21 01:59
Core Viewpoint - Tencent is restructuring its AI organization to enhance its capabilities in AI research and development, particularly focusing on large language models and foundational AI research [1][2]. Group 1: Organizational Changes - Tencent has announced the dissolution of its AI Lab, with some personnel being reassigned to the large language model department and the mixed Yuan team, reporting to Chief AI Scientist Yao Shunyu [1]. - The AI Lab, established in 2016, was part of Tencent's enterprise-level AI research, focusing on areas such as vision, language, natural processing, and machine learning [2]. - The restructuring is seen as a consolidation of Tencent's foundational AI research capabilities, with the mixed Yuan team now being the primary focus for foundational research [2]. Group 2: Strategic Developments - Tencent's recent adjustments signal a commitment to advancing its AI research, with a clear strategic plan for the next two to three quarters aimed at delivering quantifiable progress [4]. - The company has reestablished its pre-training and reinforcement learning infrastructure to improve data quality, which is expected to enhance the intelligence level of its models and accelerate overall iteration [4]. - A significant update to the mixed Yuan model, HY 3.0, is set to be launched in April, promising notable improvements over the previous version, HY 2.0, particularly in reasoning and agent capabilities [4].