[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-china-supercomputing-deepseek-v4-free-1m-context":3,"news-related-77cd56b5-87cf-4ffe-9309-a85ae5d81da9":36},{"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},"77cd56b5-87cf-4ffe-9309-a85ae5d81da9","国家超算互联网上线DeepSeek-V4限时免费服务：百万Token上下文国产大模型开放新入口","4月26日，国家超算互联网正式推出DeepSeek-V4限时免费对话服务，用户只需登录超算互联网Chat页面即可免费体验具备百万Token超长上下文的DeepSeek-V4模型实时对话。此次免费服务的推出，标志着国产开源大模型在开放获取方面迈出了重要一步。\n\nDeepSeek-V4是DeepSeek系列的开源重磅升级，在超长上下文处理能力上达到百万Token级别，这一能力对于需要处理长文档、代码库分析、多轮复杂推理的应用场景具有重要价值。从技术参数来看，V4版本在长上下文窗口下的注意力机制和推理效率上进行了专项优化，使得超长输入不再以牺牲响应速度为代价。\n\n值得注意的是，DeepSeek-V4采用了开源策略，开发者可直接在超算互联网AI社区下载模型文件，并通过众智FlagOS统一软件栈在万卡超集群系统上完成部署。这一开源路径为国内中小企业和研究机构提供了低成本获取顶级大模型能力的可能，有助于降低AI应用开发的技术门槛。\n\n从行业影响来看，国家超算互联网此举不仅是推广国产模型的技术动作，更是构建国产AI开发生态的关键布局。在当前大模型竞争日趋激烈的背景下，拥有自主可控的开源基座和低成本的获取渠道，对于整个国产AI生态的健康发展至关重要。","https:\u002F\u002F36kr.com\u002Fnewsflashes\u002F3783020652518401","5e4fd3d1-9cb4-44a6-bae5-9ffb449c05c1",[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},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"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},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"af948c2d-fc07-481d-bc2e-3b6525ceb1ec","en","China's supercomputing grid offers free DeepSeek-V4 access","On April 26, the National Supercomputing Internet officially launched the DeepSeek-V4 limited-time free chat service. Users only need to log in to the Supercomputing Internet Chat page to experience free real-time conversation with the DeepSeek-V4 model featuring million-token ultra-long context. The launch of this free service marks an important step for domestic open-source LLMs in terms of open access.\n\nDeepSeek-V4 is a major open-source upgrade in the DeepSeek series, achieving million-token level in ultra-long context processing capability — important value for application scenarios requiring long-document, code repository analysis, and multi-turn complex reasoning. From technical parameters, the V4 version has done specific optimization on attention mechanism and inference efficiency under long context windows, so ultra-long input no longer comes at the cost of response speed.\n\nNotably, DeepSeek-V4 adopts an open-source strategy. Developers can directly download the model file in the Supercomputing Internet AI community, and complete deployment on 10,000-card super-cluster systems via the FlagOS unified software stack. This open-source path provides domestic SMEs and research institutions with low-cost access to top-tier LLM capabilities, helping lower the technical threshold for AI application development.\n\nIn terms of industry impact, the National Supercomputing Internet's move isn't just a technical action to promote domestic models, but a key layout to build a domestic AI development ecosystem. Against the backdrop of increasingly fierce LLM competition, having an autonomous, controllable open-source base and low-cost access channels is crucial for the healthy development of the entire domestic AI ecosystem.","china-supercomputing-deepseek-v4-free-1m-context","2026-04-26T13:00:00Z","2026-04-26T13:05:12.543864Z","2026-08-19T02:08:40.142862Z",true,"agent",105,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"d5203c93-2022-4769-a6d2-c7765ded2b40","腾讯混元 Hunyuan-A13B 开源实测:80B 总参 \u002F 13B 激活,GQA + FP8\u002FINT4 把 MoE 推理门槛打到消费卡","tencent-hunyuan-a13b-80b-13b-gqa-angelslim-moe","2026-07-30T06:00:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"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":49,"title":50,"news_slug":51,"published_at":52},"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":54,"title":55,"news_slug":56,"published_at":57},"68072ee1-fc37-4064-ab18-09550ae72d1b","GLM-5.3-Flash 把 320B MoE 跑在国产芯片上:Flash 价位和 $0.15 API 的混合注意力栈","glm-5-3-flash-chinese-chips-hybrid-attention","2026-08-27T03:00:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"cac485ab-a429-4ebc-88c6-a1f924f978ff","AWS 开源 KeysAndValues:微调时就让模型学会“遗忘”,单张 A100 撑住 128K","aws-keysvalues-sparse-attention-finetuning","2026-08-26T05:20:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"0d8fdf45-4585-47c0-9e78-3652e318b156","Apple Intelligence 中国版落地:通义千问接管语言 AI,百度负责视觉搜索","apple-intelligence-china-qwen-baidu-2026","2026-08-25T12:00:00+00:00"]