[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-microsoft-build-2026-mai-vs-openai-10x":3,"topics-all":37,"news-related-7c78c5e3-8944-4719-8896-f351c6cad039":56},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":35,"view_count":36},"7c78c5e3-8944-4719-8896-f351c6cad039","微软 Build 2026 密集发布 MAI 系列自研模型：剑指 OpenAI，效率宣战","微软 Build 2026 密集发布 MAI 系列自研模型：剑指 OpenAI，效率宣战\n微软本周在旧金山举行的 Build 2026 开发者大会上，正式推出首批自研 AI 模型产品线，正式向 OpenAI 的霸主地位发起挑战。CEO Satya Nadella 在台上直言：\"每家公司都要从消费前沿模型，转向全面参与前沿生态系统。\"这番宣言的背景，是微软已向 OpenAI 投入 130 亿美元、向 Anthropic 投入 50 亿美元——而现在，它要自己做模型，省掉中间商。\n**三款产品，覆盖三大场景**\n第一款是 MAI-Code-1-Flash，首个代码生成模型。用户输入文字描述，即可生成应用或网站的源代码。目前已接入 GitHub Copilot AI 编程服务和 VS Code 文本编辑器，主打\"推理超高效率\"。\n第二款是 MAI-Thinking-1，推理模型，定位\"高效率、高性能、低 token 成本\"，通过 Microsoft Foundry 提供私密预览。企业用户可导入自有数据来提升推理准确率。\n第三款是一组面向 Windows PC 的小型 Aion 模型，覆盖语音识别、合成语音和图片生成三个细分能力，让端侧 AI 成为可能。\n**10 倍效率，超越 GPT-5-5**\n发布会的核心卖点是效率。微软 AI 业务 CEO Mustafa Suleyman 透露，微软在拿到麦肯锡这个大客户的真实需求后，对模型进行了针对性优化，最终在对比测试中以\"10 倍的成本效率\"超越 OpenAI 的 GPT-5-5。这里的\"成本效率\"指的是：完成同等任务消耗的 token 越少，成本越低。\nDAIGLE 在官方博客中写道：\"token 是模型读取、处理和生成数据的基本单元，token 使用量直接决定开发者的成本。\"MAI 系列的策略就是用更少的 token 做更多的事，把节省下来的成本传递给开发者。\n**大背景：IPO 竞跑**\n就在 Build 前一天，Anthropic 于 6 月 1 日秘密提交了 IPO 申请，OpenAI 也在积极筹备上市。微软此时高调宣示自研模型能力，既是向华尔街证明自己不只是\"中间商\"，也是为即将到来的 AI 平台竞争抢占话语权。\n这意味着，云厂商的 AI 战争正在从\"谁的模型调用量最大\"转向\"谁的模型成本最低、效率最高\"。","https:\u002F\u002Fwww.cnbc.com\u002F2026\u002F06\u002F02\u002Fmicrosoft-unveils-new-ai-models-lessen-reliance-on-openai-lower-costs.html","b506c01d-ef58-49cc-8ba1-351a47e7d6d1",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"7ac06d8e-b074-4147-abfc-ffaa4c6b8744","ai-efficiency",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":21,"name":22,"slug":22,"description":13,"color":13},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"0fb9088e-3462-428e-a2c3-ad9bfd82673b","en","Microsoft Build 2026 floods out MAI models to challenge OpenAI","At Build 2026 Microsoft released a dense suite of MAI series self-developed models, covering reasoning, vision, coding, and audio. The strategy is clear: reduce dependence on OpenAI, lower inference cost, and offer a Microsoft-native alternative for enterprise customers. The \"efficiency war\" framing is Microsoft's positioning — self-developed models optimized for inference cost vs OpenAI's capability-first approach.","microsoft-build-2026-mai-vs-openai-10x","2026-06-03T02:10:00Z","2026-06-03T10:06:36.909276Z","2026-08-19T02:08:40.142862Z",true,"agent","最佳盟友的姿态，密集发布了三款自研模型：MAI-Code-1-Flash、MAI-Thinking-1，以及面向 Windows PC 的小型 Aion 系列。这不是小打小闹的试水，而是一次清晰的战略宣言——微软要亲手掌握 AI 模型层，不再甘于做第三方模型的转售商。\n\n具体来看，MAI-Code-1-Flash 是微软首款代码生成模型，已直接集成到 GitHub Copilot 和 Visual Studio Code 中，主打推理超高效率。而 MAI-Thinking-1 则是一款中等规模的推理模型，微软 CEO Satya Nadella 在台上毫不客气地放话：每家公司都应该从消费前沿模型转向完全参与前沿生态系统。更让业内震动的是，微软 AI 负责人 Mustafa Suleyman 透露，经过为咨询公司麦肯锡的定制优化后，MAI 模型在性能上已超越 OpenAI 的 GPT-5.5，且成本效率提升十倍。\n\n这场发布背后有清晰的经济逻辑：微软累计向 OpenAI 投资 130 亿美元、向 Anthropic 投资 50 亿美元，每年向这两家支付的 API 调用费用惊人。自研模型可以让微软把调用成本留在自己的 Azure 基础设施上，而不是流向竞争对手。Gemini 3.5 Flash 在五月中旬发布时同样主打低成本高效，两家巨头在效率这条赛道上的贴身肉搏已经白热化。\n\n对开发者而言，这当然是好消息——选择多了，价格战最终会让使用成本下降。但对 OpenAI 和 Anthropic 而言，微软从最大的渠道商变成最直接的竞争对手，这个角色转换带来的压力远比任何一款新模型发布都更值得警惕。",178,[38,47],{"slug":39,"tag_slug":39,"title_zh":40,"title_en":41,"intro_zh":42,"intro_en":43,"id":44,"is_active":33,"created_at":45,"modified_at":46},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":48,"tag_slug":48,"title_zh":49,"title_en":50,"intro_zh":51,"intro_en":52,"id":53,"is_active":33,"created_at":54,"modified_at":55},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":57},[58,63,68,73,78,83],{"id":59,"title":60,"news_slug":61,"published_at":62},"4d436945-18e9-4d69-a4c8-c1e3e975ab33","MiniMax M3发布：稀疏注意力打通百万token上下文，开源模型编程能力逼近闭源前沿","minimax-m3-sparse-attn-million-token-msa","2026-06-04T01:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"6f4d1046-ee70-4a06-9261-2cc187c66285","12 万美元 token 把 Copilot 运行时搬进 Rust:AI 智能体包揽 43 万行移植","copilot-runtime-rust-agentic-port","2026-09-20T19:11:22+00:00",{"id":69,"title":70,"news_slug":71,"published_at":72},"b667e52f-ec7d-4ca4-8d9e-1db81e1a5616","DeepSeek论文:890字节KV缓存的三层架构账","deepseek-v41-flash-kv-cache-paper","2026-09-18T15:10:00+00:00",{"id":74,"title":75,"news_slug":76,"published_at":77},"8ebbcd9c-31ee-4baa-b395-b104bd87c8e1","Kimi K2.8 Preview 把 K3 的百万上下文下放给免费档：月之暗面的「过日子」模型登场","kimi-k2-8-preview-coding","2026-09-17T03:00:00+00:00",{"id":79,"title":80,"news_slug":81,"published_at":82},"4c4a2a9e-f69b-4985-bd42-97ab2ef4e2ac","Spark-X2.5-4B 开源:4B 跑 1M 上下文,22 项基准打 9B 级 Qwen3.5","spark-x2-5-4b-apache-open-source","2026-09-16T01:30:00+00:00",{"id":84,"title":85,"news_slug":86,"published_at":87},"01a6593b-449e-493e-ac45-33c23c9211ba","SWE-2 距 Fable 5.1 一分:2.8T 开源底座后训练,成本砍 64%","cognition-swe-2-kimi-k3-pareto","2026-09-11T21:08:02+00:00"]