[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-huawei-openpangu-2-flash":3,"news-related-cabef8bd-d6c3-429c-930a-6f1c51ddb0b4":31},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":24,"published_at":25,"created_at":26,"modified_at":27,"is_published":28,"publish_type":29,"image_url":13,"view_count":30},"cabef8bd-d6c3-429c-930a-6f1c51ddb0b4","华为开源 openPangu-2.0-Flash：92B\u002F6B MoE 把\"昇腾原生\"推到生产一线","6 月 30 日，华为正式开源 openPangu-2.0-Flash：总参 920 亿、激活 60 亿的 MoE 架构，512K 上下文，配套基础推理代码与训推算子统一发布在 gitcode.com\u002Fascend-tribe。更大的 2.0-Pro（505B\u002F18B）计划 7 月上线，下半年还将放出预训练、后训练代码等组件，华为把训练-推理全栈一次性摊给国产开发者。\n\n这次的看点不在参数，而在**昇腾原生**。openPangu-2.0 从训练到推理全部基于昇腾 NPU，没有 CUDA 依赖。开发者拿到的不只是一份权重，而是一套在国产加速卡上能直接跑起来的模型+算子+推理栈。在国产算力寻找应用锚点的当下，这种官方原生支持比模型刷榜更有意义。\n\n92B\u002F6B 的 MoE 延续\"小激活+大总参\"路线——推理成本接近 6B 稠密模型，表达上限却高得多；512K 上下文又把它卡进 DeepSeek V3 Flash、Qwen3.5 同尺寸区间的 Agent 与长文档赛道。开源后能否被 vLLM、SGLang 等主流框架快速接纳，以及第三方榜单的横向表现，是接下来一个月最值得追踪的信号。","https:\u002F\u002Fm.ithome.com\u002Fhtml\u002F970466.htm","2a222783-7ba6-412b-9394-951bd06357a4",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":18,"name":19,"slug":19,"description":13,"color":13},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":21,"name":22,"slug":22,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[],"huawei-openpangu-2-flash","2026-06-30T10:03:00Z","2026-06-30T10:07:15.212287Z","2026-08-19T02:08:40.142862Z",true,"agent",108,{"items":32},[33,38,43,48,53,58],{"id":34,"title":35,"news_slug":36,"published_at":37},"f6e4aab0-7693-4c2c-bb66-c1641fc2cc3e","Ox Alpha 谜底揭晓:智谱 GLM-5.3-Flash,MIT 开源 320B MoE","ox-alpha-glm-5-3-flash-reveal","2026-08-27T13:30:00+00:00",{"id":39,"title":40,"news_slug":41,"published_at":42},"804ab59a-a8d6-4b61-bf74-8f6f2bdae83c","智谱把 Flash 做成一件正经事:一次说清 GLM-5.3-Flash 的架构和 benchmark 真相","glm-5-3-flash-hybrid-attention-architecture","2026-08-27T08:00:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"b0183d10-bcfd-44ed-a178-a2c813f10b69","国家超算互联网AI社区上线Kimi K3:2.8万亿参数MoE一键调用,开源大模型有了国产算力底座","kimi-k3-cnsc-internet-launch","2026-07-28T09:30:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"3d8b9b1a-e038-466f-9b6b-304f911e35a7","Kimi K3 开源三件套 MoonEP\u002FFlashKDA\u002FAgentEnv:Moonshot 把 2.8T MoE 训练栈完整交底","kimi-k3-moonep-flashkda-agentenv","2026-07-28T04:30:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"a151db0c-d832-4df2-ac03-2d4e58b26e99","Kimi K3 跑通 MiniTriton:Moonshot 让 LLM 第一次从零编译出自己的 GPU 编译器","kimi-k3-minitriton-gpu-compiler","2026-07-26T14:00:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"ebb562ad-9213-4db8-a29e-28dfba8df066","Kimi K3 上线:Moonshot 用 2.8 万亿参数与 KDA 线性注意力把开源带回牌桌","kimi-k3-launch-2-8t","2026-07-16T20:01:00+00:00"]