[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-apple-m7-end-side-llm":3,"topics-all":36,"news-related-44190b56-9916-48cb-b08a-d041ac7b857c":55},{"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":13,"view_count":35},"44190b56-9916-48cb-b08a-d041ac7b857c","苹果跳过 M6 Pro\u002FMax,提前半年发 M7:这是端侧大模型算力的最强信号","据马克·古尔曼报道,苹果打破\"M 芯片节奏\":今年秋季照常推 M6,但 M6 Pro\u002FMax 不会按惯例春季发布。原定 2027 下半年才出的 M7 提前半年,MacBook Pro、Mac Studio 用户最快 2027 上半年拿到 M7。\n\nM7 被定位为\"AI 导向\",代号已曝光:M7(Delos\u002FH19G)、M7 Pro(H19S)、M7 Max(H19C),M7 Ultra(H19D)要等到 2028。这是工作站被往前推的算力节奏。\n\n工程圈最该关注数字:Gurman 透露 M7 基础版内存带宽将达 240GB\u002Fs,比 M6(200GB\u002Fs)高 20%,比 M5(153GB\u002Fs)高 57%。这意味着 16GB 统一内存跑 30B 稠密模型首字延迟几乎打对折,是 7B\u002F13B 本地推理告别卡顿的关键门槛。Neural Engine 升级在 M6 释放,M7 把\"AI 偏向\"推到极致。\n\n行业层面是\"反摩尔\"算力加速:NVIDIA RTX Spark 即将杀入 Windows AI PC,AMD、Intel、高通都在端侧堆参数。当 M7 Pro\u002FMax 把\"240GB\u002Fs 带宽 + 128GB+ 统一内存\"做成标配,本地跑 70B 模型与 Agent 长链推理就不再是 macOS 的特权,而是它的主战场。","https:\u002F\u002F36kr.com\u002Fnewsflashes\u002F3893331854490120","5e4fd3d1-9cb4-44a6-bae5-9ffb449c05c1",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"fca9258a-9430-455a-b95d-b9fae5e373a8","ai-inference",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",{"id":18,"name":19,"slug":19,"description":13,"color":13},"e0d31e94-ce47-4c8f-831c-d3d2926d42f3","hardware",{"id":21,"name":22,"slug":22,"description":13,"color":13},"95d7995a-fddb-47ba-b8e6-e976ac65414b","strategy",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"8048121b-6fb1-40cc-91ba-aaac031e4f94","en","Apple skips M6 Pro\u002FMax, ships M7 early: an on-device AI signal","According to Mark Gurman, Apple has broken the \"M chip rhythm\": the M6 will be released on schedule this fall, but the M6 Pro\u002FMax won't appear in the customary spring window. The M7, originally planned for the second half of 2027, has been moved up by six months, with MacBook Pro and Mac Studio users getting the M7 as early as the first half of 2027. The M7 is positioned as \"AI-oriented\", with codenames already revealed: M7 (Delos\u002FH19G), M7 Pro (H19S), M7 Max (H19C); the M7 Ultra (H19D) won't arrive until 2028. This is the compute rhythm of a workstation being pushed forward. The number the engineering community should pay most attention to: Gurman revealed that the M7 base will have 240GB\u002Fs memory bandwidth — 20% more than M6 (200GB\u002Fs) and 57% more than M5 (153GB\u002Fs). This means a 16GB unified-memory run of a 30B dense model will nearly halve first-token latency, which is the key threshold for 7B\u002F13B local inference to say goodbye to stuttering. The Neural Engine upgrade ships with M6; the M7 pushes the \"AI bias\" to the extreme. On the industry side it's \"anti-Moore\" compute acceleration: NVIDIA RTX Spark is about to invade the Windows AI PC, and AMD, Intel, and Qualcomm are all piling parameters on the device. When the M7 Pro\u002FMax makes \"240GB\u002Fs bandwidth + 128GB+ unified memory\" the default, running a 70B model locally with Agent long-chain reasoning is no longer a macOS privilege — it's macOS's main battlefield.","apple-m7-end-side-llm","2026-07-13T08:12:17Z","2026-07-13T08:22:24.656384Z","2026-08-19T02:08:40.142862Z",true,"agent",254,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"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":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"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":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"509d88cb-d9df-4dab-8863-3832fc52897a","英伟达每颗 GPU 藏 10-40 个 RISC-V 核心：Falcon 退役、GSP 上位","nvidia-risc-v-gpu-falcon-gsp","2026-09-21T03:00:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"b48106fa-b9c7-4317-a40b-61e4219a698a","IBM Z 双架构处理器:把 Arm 推进大型机,内置 AI 推理","ibm-z-dual-architecture-arm-ai-inference","2026-09-03T03:00:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"5d452086-ecb7-494d-82b9-d962664aa243","IBM 把 Arm 核塞进 Z 大型机:Hot Chips 2026 公布业界首款双指令集处理器","ibm-z-arm-dual-isa-hot-chips-aug-2026","2026-08-29T06:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"fb1cbe25-8b85-41ec-b619-9a27b405ec34","AMD 收购 Taalas:把 AI 模型权重「刻进硅片」的推理新打法","amd-acquires-taalas-inference-chip","2026-08-19T01:00:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"45375854-7739-4dd1-bc6a-30db4474652a","Taalas HC2:把单片参数拉到 200 亿,「模型刻进硅片」的第二章","taalas-hc2-20b-mxfp4-50-chips-1t-amd","2026-08-19T00:00:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"dfdc3216-52aa-4a78-9bf5-859affc37d17","AMD 收下 Taalas：把模型权重刻进芯片，推理的内存墙还剩多少？","amd-acquires-taalas-msic-etched-weights","2026-08-11T02:00:00+00:00"]