[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fhSx6c2VejAy_u4JHq2Oy8IdbZt6sRMkJO1PqRrycq00":3},{"code":4,"msg":5,"data":6},200,"操作成功",{"id":7,"title":8,"content":9,"digest":10,"source":10,"coverPath":11,"thumbsCoverPath":12,"isTop":13,"isShow":14,"baseClick":13,"clickCount":15,"createTime":16,"typeId":17,"isNewest":18,"newsInfoTypeRespVo":19,"voiceUrl":22,"voiceSize":23,"taskId":24,"releaseTime":25,"titleEn":26,"contentEn":27,"voiceUrlEn":28,"taskIdEn":29,"voiceSizeEn":30},1539,"re:Invent透视：亚马逊云科技的Agentic布局","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">不久前，亚马逊云科技首席执行官Matt Garman在re:Invent大会上描绘了一个由“数十亿AI Agent”驱动的未来，他直言“Agent的出现使我们在AI轨迹上发生了变化——从一个技术奇迹的时代，转向真正获得价值的时代。”\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">纵观此次re:Invent大会上的密集发布，可以看出亚马逊云科技在Agentic AI时代的布局：通过贯穿从底层基础设施、推理平台、Agent构建平台到Agent应用的完整技术栈，系统性降低企业构建和部署生产级AI的门槛。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F655dea51ddbb41b58e4b6609c4825777.png\" width=\"749\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"color: rgb(255, 153, 0); font-size: 18px;\">01\u003C\u002Fspan>\u003Cspan style=\"color: rgb(255, 255, 255); font-size: 18px;\">  \u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">AI基础设施：自研芯片夯实算力基石\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">在Agentic时代，工作负载的复杂性要求，对底层算力提出了前所未有的挑战。亚马逊云科技的应对之策是“两条腿走路”：在算力层面追求极致的性价比，在部署模式上提供前所未有的灵活性。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">首先，是自研芯片的持续突破。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">本次大会，亚马逊云科技推出最新一代自研训练芯片Amazon EC2 Trainium3 UltraServers，相比前代，其计算性能提升4.4倍，而更为关键的能效指标：每兆瓦功耗处理的AI Token数量提升了5倍。这直接回应了企业对于大模型训练成本的关切。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">更引人注目的是对下一代Trainium4的预告，其承诺将在FP4计算性能上再提升6倍。这种快速的迭代节奏，彰显了亚马逊云科技通过自研芯片定义AI算力的决心。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">同时，亚马逊云科技推出迄今性能最强、能效最高的自研处理器Amazon Graviton5。目前，亚马逊云科技新增CPU容量中连续第三年有超半数由Graviton驱动，EC2头部1000家客户中已有98%受益于其显著的性价比优势，包括Adobe、Airbnb、Atlassian、Epic Games、Formula 1、Pinterest、SAP、Siemens、Snowflake与Synopsys等。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F55cb96d505b44732a3d22ed289a4ca75.png\" width=\"781\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">其次，是开创性的混合部署模式。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">新推出的 Amazon AI Factory 服务允许企业在自己的数据中心内部署由亚马逊云科技提供的专用AI基础设施，最新的NVIDIA加速计算平台、Trainium芯片、AI服务以及亚马逊云科技高速低延迟的网络。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Amazon AI Factory由亚马逊云科技负责集成基础设施的部署与管理，避免自建过程中的复杂性。本质上是亚马逊云科技将其“AI基础设施能力栈”进行了空间解耦，让算力去适应客户数据的位置，而非相反。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"color: rgb(255, 153, 0); font-size: 18px;\">02\u003C\u002Fspan>\u003Cspan style=\"color: rgb(255, 255, 255); font-size: 18px;\"> \u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">推理平台：多元模型与自研模型迭代\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">企业业务场景的多样性，决定了没有万能的AI大模型。亚马逊云科技在推理平台层面推行“多元主义”生态战略，同时以自研Nova系列模型不断提升性价比。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">其全托管服务 Amazon Bedrock 是这一战略的承载核心。本次大会Amazon Bedrock一次性新增了18个开放权重模型。包含来自Google的Gemma 3、来自Moonshot AI的Kimi K2 Thinking、来自MiniMax的MiniMax M2，以及NVIDIA的Nemotron和OpenAI的GPT OSS Safeguard等多款热门模型。这种“模型开放市场”的策略赋予了企业根据需求选择最佳工具的灵活性。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002Fa602dd230e824a7d8af828df958d3419.png\" width=\"788\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">同时，亚马逊云科技推出了自研的 Amazon Nova 2 模型家族，旨在定义AI推理的“性价比”新标准。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">其中，Nova 2 Lite是面向日常工作负载的快速、经济型推理模型；\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Nova 2 Pro是亚马逊最智能的推理模型，能够处理文本、图像、视频和语音输入，并生成文本输出，也被称作“教师模型”；\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Nova 2 Omni是亚马逊的端到端语音模型，该模型支持更多语言和富有表现力的音色，并提供高达100万 tokens的上下文窗口，能够在语音与文本之间无缝切换。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"color: rgb(255, 153, 0); font-size: 18px;\">03\u003C\u002Fspan>\u003Cspan style=\"color: rgb(255, 255, 255); font-size: 18px;\"> \u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">数据底座：海量存储与向量搜索结合\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">“让模型去了解您的数据是至关重要的，因为数据是您区别于竞争对手的关键。”亚马逊云科技大中华区产品部总经理陈晓建表示。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F983505b1de1a4b48a888535926542ed9.png\" width=\"791\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">此次大会，云存储服务Amazon S3发布一系列重磅更新，包括Amazon S3 Vectors正式可用并实现每索引20亿向量的规模突破、Amazon S3对象最大容量从5TB扩展至50TB、Amazon S3 Batch Operations处理速度提升10倍等。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">通过将Amazon S3 Vectors的海量存储能力与向量搜索Amazon OpenSearch Service的实时查询能力结合使用，足以满足丰富场景下的数据存储和检索需求。例如Adobe或国内的一些智能摄像头客户，原本就在使用Amazon S3存储海量视频，现在还可以通过Amazon OpenSearch Service对视频和音频数据进行自然语言搜索。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"color: rgb(255, 153, 0); font-size: 18px;\">04\u003C\u002Fspan>\u003Cspan style=\"color: rgb(255, 255, 255); font-size: 18px;\"> \u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">Agent构建工具：跨越从“试点”到“生产”的鸿沟\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">今年亚马逊云科技推出了AgentCore，贯穿从构建、部署到运营Agent的完整开发生命周期，旨在帮助企业快速将Agent从试点阶段推进到生产环境。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AgentCore不是单一工具，而是一个模块化的“生产就绪”组件库，提供构建企业级Agent所需的一切后台能力：安全的运行时环境、记忆管理、身份集成、可观测性等。本次大会为其新增的三项新功能，直指生产部署的核心痛点：\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Policy in Amazon Bedrock AgentCore，用于帮助企业为Agent的操作设定清晰的边界。团队可以使用自然语言，通过定义Agent可访问的工具和数据、可执行的操作以及适用条件，为其划定具体范围。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AgentCore Evaluations旨在评估AI Agent的质量，它会持续采样Agent的实时交互数据，并根据正确性、实用性、安全性等预置标准分析Agent行为。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AgentCore Memory强化了AI Agent的记忆能力。能够从过往经验中学习，并在后续交互中优化决策。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这些新功能共同将Agent开发从“手工作坊”模式，推进到了标准化、可管控、可评估的“工业化”平台阶段。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">除了提供工具，亚马逊云科技更通过发布Kiro、Amazon Security Agent、Amazon DevOps Agent三款“前沿Agent”，为业界展示了AI Agent的能力范式和颠覆性潜力。其中，Kiro是一个高度智能化的编程助手。它采用Spec-Driven Programming（规格驱动编程），通过人的指令逐步分解并自动化完成任务，同时能与人交互并接受建议，最终达成目标。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"font-size: 18px;\">结语\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Agentic时代的竞争，本质上是全栈技术能力与平台化生态的竞争。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">透视re:Invent 2025，亚马逊云科技正通过其全栈AI能力，加速“数十亿AI Agent”时代的到来，帮全球企业跨越从AI演示到商业价值的鸿沟。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">【新闻来源】msn \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1STY68?ocid=msedgntphdr&cvid=694b47053fd04459bef635d5d2afa751&ei=13\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（本网转发此文章，旨在为读者提供更多的信息资讯，所涉内容不构成投资、消费建议。文章事实如有疑问，请与有关方核实，文章观点非本网观点，仅供读者参考。）\u003C\u002Fspan>\u003C\u002Fp>","","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002Fa0b3fe2176b845dca3fd2776758f2ac7\u002FAI领域.jpg","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002Fthumbs\u002Fa0b3fe2176b845dca3fd2776758f2ac7\u002FAI领域.jpg",0,1,63,"2025-12-25 09:38",2,false,{"id":17,"name":20,"enName":21},"芯位视野","Xinwei Vision","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3Ad357c0ba-9b3f-4864-9252-d4df52836585%3A0.wav?Expires=1766630772&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=b1tZsQE1skbNvJBQGA9Codu23vw%3D",12597784,"d357c0ba-9b3f-4864-9252-d4df52836585","2025-12-25 09:26","re:Invent Perspective: Amazon Web Services' Agentic Layout","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">Not long ago, the CEO of Amazon Web Services, Matt Garman, painted a future driven by \"billions of AI Agents\" at the re:Invent conference. He stated directly, \"The emergence of Agents has changed our trajectory in AI - from an era of technological miracles to an era of truly obtaining value.\"\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Looking at the dense releases during this re:Invent conference, it is clear that Amazon Web Services is laying out its strategy for the Agentic AI era: through a complete technology stack that spans from underlying infrastructure, reasoning platforms, Agent building platforms to Agent applications, systematically lowering the barriers for enterprises to build and deploy production-grade AI.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F655dea51ddbb41b58e4b6609c4825777.png\" width=\"749\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"color: rgb(255, 153, 0); font-size: 18px;\">01\u003C\u002Fspan>\u003Cspan style=\"color: rgb(255, 255, 255); font-size: 18px;\"> \u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">AI Infrastructure: Self-developed Chips Lay the Foundation for Computing Power\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In the Agentic era, the complexity of workloads poses unprecedented challenges for underlying computing power. The response of Amazon Web Services is \"walking on two legs\": pursuing extreme cost-effectiveness in computing power and providing unprecedented flexibility in deployment models.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Firstly, continuous breakthroughs in self-developed chips.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">At this conference, Amazon Web Services introduced the latest generation of self-developed training chips, Amazon EC2 Trainium3 UltraServers. Compared to the previous generation, its computing performance has been improved by 4.4 times, while more importantly, the energy efficiency metric: the number of AI Tokens processed per megawatt of power has increased by 5 times. This directly addresses enterprises' concerns about the costs of large model training.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">More notably, the announcement of the next-generation Trainium4, which promises to improve FP4 computing performance by 6 times. This rapid iteration pace demonstrates Amazon Web Services' determination to define AI computing power through self-developed chips.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">At the same time, Amazon Web Services launched the most powerful and energy-efficient self-developed processor, Amazon Graviton5. Currently, over half of the newly added CPU capacity by Amazon Web Services has been driven by Graviton for three consecutive years. Among the top 1,000 EC2 customers, 98% have benefited from its significant cost-effectiveness advantage, including Adobe, Airbnb, Atlassian, Epic Games, Formula 1, Pinterest, SAP, Siemens, Snowflake, and Synopsys.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F55cb96d505b44732a3d22ed289a4ca75.png\" width=\"781\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Secondly, innovative hybrid deployment models.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The newly launched Amazon AI Factory service allows enterprises to deploy dedicated AI infrastructure provided by Amazon Web Services within their own data centers, including the latest NVIDIA accelerated computing platform, Trainium chips, AI services, and Amazon Web Services' high-speed, low-latency network.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Amazon AI Factory is responsible for the integration and management of the infrastructure by Amazon Web Services, avoiding the complexity of self-building. Essentially, it spatially decouples Amazon Web Services' \"AI infrastructure capability stack\", allowing computing power to adapt to the location of customer data, rather than the opposite.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"color: rgb(255, 153, 0); font-size: 18px;\">02\u003C\u002Fspan>\u003Cspan style=\"color: rgb(255, 255, 255); font-size: 18px;\"> \u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">Reasoning Platform: Diverse Models and Self-developed Model Iterations\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The diversity of enterprise business scenarios determines that there is no universal AI large model. Amazon Web Services implements a \"pluralism\" ecological strategy at the reasoning platform level, while continuously improving cost-effectiveness with its self-developed Nova series models.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Its fully managed service, Amazon Bedrock, is the core of this strategy. At this conference, Amazon Bedrock added 18 open-weight models at once. These include Gemma 3 from Google, Kimi K2 Thinking from Moonshot AI, MiniMax M2 from MiniMax, Nemotron from NVIDIA, and GPT OSS Safeguard from OpenAI, among other popular models. This \"model open market\" strategy gives enterprises the flexibility to choose the best tools according to their needs.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002Fa602dd230e824a7d8af828df958d3419.png\" width=\"788\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">At the same time, Amazon Web Services launched its self-developed Amazon Nova 2 model family, aiming to define a new standard for cost-effectiveness in AI reasoning.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Among them, Nova 2 Lite is a fast and economical reasoning model for daily workloads;\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Nova 2 Pro is Amazon's smartest reasoning model, capable of handling text, image, video, and voice inputs and generating text outputs, also known as the \"teacher model\";\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Nova 2 Omni is Amazon's end-to-end voice model, supporting more languages and expressive tones, and provides a context window of up to 1 million tokens, enabling seamless switching between voice and text.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"color: rgb(255, 153, 0); font-size: 18px;\">03\u003C\u002Fspan>\u003Cspan style=\"color: rgb(255, 255, 255); font-size: 18px;\"> \u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">Data Base: Combining Massive Storage and Vector Search\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">\"It is crucial for the model to understand your data because data is the key differentiator from your competitors,\" said Chen Xiaojian, General Manager of the Product Department of Amazon Web Services Greater China.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F983505b1de1a4b48a888535926542ed9.png\" width=\"791\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">At this conference, the cloud storage service Amazon S3 released a series of major updates, including the official availability of Amazon S3 Vectors and the breakthrough of 2 billion vectors per index, the maximum capacity of Amazon S3 objects was expanded from 5TB to 50TB, and the processing speed of Amazon S3 Batch Operations was increased by 10 times.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">By combining the massive storage capabilities of Amazon S3 Vectors with the real-time query capabilities of the vector search Amazon OpenSearch Service, it can meet the data storage and retrieval needs of various scenarios. For example, Adobe or some domestic intelligent camera customers, who have been using Amazon S3 to store large volumes of video, can now use Amazon OpenSearch Service to perform natural language searches on video and audio data.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"color: rgb(255, 153, 0); font-size: 18px;\">04\u003C\u002Fspan>\u003Cspan style=\"color: rgb(255, 255, 255); font-size: 18px;\"> \u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">Agent Building Tools: Crossing the Chasm from \"Pilot\" to \"Production\"\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">This year, Amazon Web Services launched AgentCore, covering the entire development lifecycle from building, deploying to operating Agents, aiming to help enterprises quickly move Agents from pilot stages to production environments.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AgentCore is not a single tool, but a modular \"production-ready\" component library, providing all the backend capabilities needed to build enterprise-level Agents: secure runtime environment, memory management, identity integration, observability, etc. The three new features added at this conference directly address the core pain points of production deployment:\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Policy in Amazon Bedrock AgentCore is used to help enterprises set clear boundaries for Agent operations. Teams can use natural language to define the tools and data that the Agent can access, the operations that can be performed, and the applicable conditions, thereby setting specific ranges.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AgentCore Evaluations aim to evaluate the quality of AI Agents. It continuously samples real-time interaction data of the Agent and analyzes the Agent's behavior based on pre-set criteria such as correctness, practicality, and security.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AgentCore Memory enhances the memory capabilities of AI Agents. It can learn from past experiences and optimize decisions in subsequent interactions.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">These new features collectively advance Agent development from a \"craft workshop\" model to a \"industrialized\" platform stage that is standardized, controllable, and assessable.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In addition to providing tools, Amazon Web Services has also demonstrated the capability paradigm and disruptive potential of AI Agents by launching three \"frontier Agents\": Kiro, Amazon Security Agent, and Amazon DevOps Agent. Among them, Kiro is a highly intelligent programming assistant. It adopts Spec-Driven Programming (specification-driven programming), breaking down and automating tasks step by step according to human instructions, while being able to interact with people and accept suggestions, ultimately achieving the goal.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Ch1>\u003Cspan style=\"font-size: 18px;\">Conclusion\u003C\u002Fspan>\u003C\u002Fh1>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Competition in the Agentic era essentially involves full-stack technological capabilities and platform-based ecological competition.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Perspective on re:Invent 2025, Amazon Web Services is accelerating the arrival of the \"billions of AI Agents\" era through its full-stack AI capabilities, helping global enterprises bridge the gap from AI demonstrations to commercial value.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">[News Source] msn \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1STY68?ocid=msedgntphdr&cvid=694b47053fd04459bef635d5d2afa751&ei=13\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（This article is reprinted by this website to provide readers with more information and news. The content involved does not constitute investment or consumption advice. If there are any doubts about the facts of the article, please verify with the relevant parties. The views expressed in the article are not the views of this website, and are for reference only.)\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A59d5514a-f917-4124-bc3e-d0acc73eb47e%3A0.wav?Expires=1774838437&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=njzhMbtVs3sz7S8KeTAyYfKTv5U%3D","59d5514a-f917-4124-bc3e-d0acc73eb47e",15516654]