[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-google-deepmind-reshuffle-gemini-speed":3,"news-related-7b9cdf6e-5ef0-4ece-ab6c-e8cec1b02397":38},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"7b9cdf6e-5ef0-4ece-ab6c-e8cec1b02397","Google 重组 DeepMind 领导层,Gemini 研发提速应对 Anthropic 与 OpenAI 竞争","据路透社报道,Google 本月初重组了 DeepMind 的领导层,目的是加速 Gemini 模型研发以追赶 Anthropic 和 OpenAI。Gemini 去年 11 月一度领先,但竞争对手的新模型让其再次陷入追赶,新版本也因编程能力不达标而被推迟发布。","Solidot 转载路透社报道,Google 在 2026 年 8 月初完成了对 DeepMind 领导层的重组。这次调整的核心信号很直接:Google 内部认为 Gemini 的研发节奏已经被 Anthropic 与 OpenAI 甩开,需要在组织结构上动手脚来提速。\n\n事情的起点是 Gemini 的发布节奏。Google 的 Gemini 模型曾在 2025 年 11 月短暂领先对手,但随后 Claude 与 GPT 系列陆续放出新版,让 Google 又回到了追赶者位置。这一次被推迟的新一代 Gemini 模型,卡点正是编程能力:内部 benchmark 显示它在 coding 维度仍然落后于头部竞品,所以 Google 选择了延期发布,而不是带着明显短板上市。\n\n组织层面的动作可以拆成两层来看。第一层是削弱 DeepMind 的自主权,把更多算力、数据和工程资源集中投向 Gemini 项目本身,这意味着 Alphabet 体系下原本分散的多条研究支线被收窄。联合创始人 Sergey Brin 在今年 4 月的一次内部讲话里强调了加快节奏的紧迫性,这次重组可以被视作那次讲话的组织落地。第二层则是执行层面的信号:Google 不再愿意等一个「全明星」版本,而是接受分阶段、可发布、有缺口的迭代,这与 Anthropic 一年多来推 Claude 的方式越来越像。\n\n对行业来说,这次重组更像是一个拐点信号,而不是单纯的人事新闻。前沿实验室之间的领先窗口已经从「半年」缩短到「几次发布」。当一家公司为了追平对手而愿意动到核心研究部门的组织结构,说明单纯靠堆算力和扩团队已经不够,决策链路本身成了瓶颈。Google 把 Gemini 从「DeepMind 的明星项目」升级为「全公司的资源汇集点」,实际上是把研究优先级让位给了产品优先级。\n\n对国内大模型团队来说,这同样值得注意。当头部公司愿意为模型节奏让渡研究自由度时,OpenAI、Anthropic 与 Google 之间的差距收敛速度会比大多数人的直觉更快。下半年的看点,在于 Gemini 新版本能否在 coding 与 agent 这两个落后项上把分追回来,以及它会不会为此牺牲一部分长尾能力。换句话说,Google 这次押注的并不是模型本身,而是组织能否跑得更快。","https:\u002F\u002Fwww.solidot.org\u002Fstory?sid=85082","d59894d3-308e-4fd8-8865-86dc1eeac4a2",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"a9524a82-a7c5-4daa-bb4b-a7ee77bb0b94","gemini",{"id":19,"name":20,"slug":20,"description":14,"color":14},"8cf7490f-2449-4ba7-be19-61befa0d92b4","google",{"id":22,"name":23,"slug":23,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"a48bcd01-2aa3-4bc9-a751-cc13d70372cb","en","Google Restructures DeepMind Leadership to Accelerate Gemini Against Anthropic and OpenAI","According to a Reuters report republished by Solidot, Google restructured DeepMind's leadership at the start of August to accelerate Gemini development and catch up with Anthropic and OpenAI. Gemini briefly led in November 2025, but rival model releases pushed it back into catch-up mode, and the latest version was delayed after internal coding benchmarks fell short.","According to a Reuters report republished by Solidot, Google completed a leadership reshuffle at DeepMind in early August 2026. The core signal is direct: internally, Google believes the Gemini development pace has fallen behind Anthropic and OpenAI, and structural changes are needed to accelerate.\n\nThe starting point is Gemini's release cadence. Google's Gemini model briefly led rivals in November 2025, but successive Claude and GPT releases pushed Google back into catch-up mode. The newest Gemini version was delayed specifically because its coding capability did not clear the internal bar. Internal benchmarks showed it still trailing top competitors in coding, so Google chose to postpone the release rather than ship a model with a visible weak spot.\n\nThe organizational changes can be read on two levels. The first is a reduction of DeepMind's autonomy, with more compute, data and product engineering resources concentrated on the Gemini project itself. That means several research threads inside Alphabet have been narrowed. Co-founder Sergey Brin emphasized the urgency of moving faster in an internal talk in April this year, and this reshuffle can be seen as the organizational landing of that message. The second is an execution-level signal: Google is no longer willing to wait for a single \"all-star\" release and is accepting staged, shippable, imperfect iterations, much closer to how Anthropic has been shipping Claude for over a year.\n\nFor the industry, this reshuffle reads more like an inflection signal than a routine personnel story. The lead window between frontier labs has compressed from \"half a year\" to \"a few releases.\" When a company is willing to restructure its core research organization just to close the gap with competitors, it means that stacking more compute and headcount is no longer enough, the decision chain itself has become the bottleneck. By upgrading Gemini from \"DeepMind's marquee project\" to \"the company's resource convergence point,\" Google is effectively letting product priority override research priority.\n\nChinese large-model teams should also pay attention. When the leading labs are willing to trade research freedom for shipping speed, the convergence rate between OpenAI, Anthropic and Google will be faster than most people expect. The question for the second half of the year is whether the next Gemini release can close the gap in coding and agent capabilities, and whether it will sacrifice some long-tail capability to do so. In other words, Google's bet here is not on the model itself, but on whether the organization can run faster.","google-deepmind-reshuffle-gemini-speed","2026-08-25T07:00:00Z","2026-08-25T05:02:43.137872Z","2026-08-25T05:02:43.137880Z",true,"agent",57,{"items":39},[40,45,50,55,60,65],{"id":41,"title":42,"news_slug":43,"published_at":44},"bcedeb8e-e5eb-4bbc-98b8-ea12f869055f","Google 收编 DeepMind：25 年最大 AI 重组","google-deepmind-centralization-gemini-catchup","2026-08-14T08:00:00+00:00",{"id":46,"title":47,"news_slug":48,"published_at":49},"4bd8e8bd-7066-4ab7-bd97-e24ea3921395","Gemini 因编程落后推迟两月:Brin 4 月督促背后,Google 把研发「收回到一个人」手里的组织账本","google-gemini-coding-behind-deepmind-reshuffle","2026-08-14T03:30:00+00:00",{"id":51,"title":52,"news_slug":53,"published_at":54},"8b6c20ec-7222-48cf-af2c-ac97466a2b0a","Gemini 月活破 10 亿:Google 第一次把 AI 助手做成自家「最快十亿用户产品」","gemini-app-1b-monthly-users","2026-08-12T03:00:00+00:00",{"id":56,"title":57,"news_slug":58,"published_at":59},"34edaffc-6b5c-4df1-9e2f-d864cada6063","Gemini 走进 K-12 课堂：Google 把「上下文」塞进每个作业","gemini-classroom-k12-contextualized-prompts","2026-08-07T02:00:00+00:00",{"id":61,"title":62,"news_slug":63,"published_at":64},"df25ef84-5f0b-477b-9940-a0da477d8169","DeepMind 新主帅接棒：Hassabis 退任，Gemini 4 成 Google 筹码","deepmind-kavukcuoglu-gemini-4-reshuffle","2026-08-07T00:00:00+00:00",{"id":66,"title":67,"news_slug":68,"published_at":69},"d8a15300-7bba-40d9-a62c-436a2f3a68bc","Google ATLAS 报告:1500 万次 Gemini 交互告诉我们什么?","google-gemini-atlas-work-adoption-data","2026-07-30T03:30:00+00:00"]