[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-cnnic-china-genai-users-700m-2026":3,"topics-all":38,"news-related-9c9e09c8-dad6-47b6-a6d5-02813b939de6":57},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"9c9e09c8-dad6-47b6-a6d5-02813b939de6","CNNIC:2026 上半年中国生成式 AI 用户突破 7 亿,普及率超 50%","中国互联网络信息中心发布《生成式人工智能应用发展报告(2026)》,截至 2026 年上半年,中国生成式 AI 用户规模突破 7 亿,普及率超过 50%。76% 用户用 AI 回答问题,47.8% 用 AI 处理图像视频,38.7% 网民近半年买过智能硬件。报告强调本土团队在全球主流大模型调用榜单前六全部来自中国。","## 报告数据:中国生成式 AI 半年新增到什么规模\n\n中国互联网络信息中心(CNNIC)发布的《生成式人工智能应用发展报告(2026)》给出一个全景数字:截至 2026 年上半年,中国生成式 AI 用户规模突破 7 亿,普及率超过 50%。这意味着每隔两个中国网民就有一个在用生成式 AI 工具,而半年前这个数字还没过半数——CNNIC 把这条曲线画得很陡。\n\n报告把用户行为拆开看,76.0% 的用户表示会让生成式 AI 回答问题——问答依然是绝对主入口。真正的增量来自多模态:47.8% 用 AI 处理图像和视频,37.6% 拿 AI 处理文本生成,32.5% 把它写进工作总结、会议纪要和 PPT 的工作流。AI 已经渗进了办公场景的三个高频环节。\n\n## 硬件入口:穿戴和 3C 数码是渗透主战场\n\n报告里另一组常被忽略的数字——38.7% 的网民过去半年在网上买过智能硬件设备。其中可穿戴设备占 20.2%,智能手机、平板电脑等 3C 数码产品占 18.2%。生成式 AI 的\"软硬协同\"路径不是从 AR 眼镜这种新形态起步,而是从已经成熟智能手表和手机平板先吃一波红利。\n\n把这两组数字拼在一起,逻辑就清楚了:可穿戴和 3C 数码是硬件入口,生成式 AI 是软件入口,两端都在加速。一个用户可能买了新的智能手表,又在手机端频繁调用 AI 助手——两端都在 CNNIC 报告里拿到了官方背书。\n\n## 中国厂商在\"调用榜单前六\"全部上榜\n\n报告最敏感的一组数据:全球主流大模型调用榜单上,排名前六的模型全部来自中国团队。报告点名了深度求索(DeepSeek)和月之暗面(Moonshot Kimi),称本土企业先后发布多个万亿级参数开源大模型。\n\n这条信息的口径需要谨慎。\"调用榜单\"具体指哪家机构、哪份榜单,CNNIC 在转载版本里没有点名。但如果是真实数据,它的指向是明确的:过去两年中国大模型在推理调用量、Token 消耗、开发者接入数这些工程指标上,已经在全球前列站稳脚跟。开源权重模型在这里扮演角色,和闭源 API 完全不同——它们提供低成本、本地化、可微调的底座,把调用量推到第一梯队。\n\n## 数据之外的几个判断\n\n第一,7 亿用户、50% 普及率这两个数字来自 CNNIC 抽样调查,不是运营商或支付平台级别的硬数据。它的强项是覆盖广,弱项是频率定义宽松——\"使用过\"和\"日常使用\"是两回事。\"突破 7 亿\"更准确读法是使用过生成式 AI 的中国网民超过 7 亿。\n\n第二,76% 问答、47.8% 多模态、37.6% 文本、32.5% 办公——这组比例在同一个用户身上可以叠加,加总远超 100%。它说明 AI 工具不是单点渗透,而是场景叠加。\n\n第三,前六全部来自中国这条,跟开源生态和价格战直接相关。DeepSeek-V3、Kimi K2、Qwen3 系列等开源权重模型把推理单价压到了闭源 API 的零头,开发者用脚投票——但\"全球调用前六\"是否把内部测试流量也算进去,报告没有交代。\n\n## 所以呢\n\nCNNIC 这份报告最值得读的不是 7 亿这个数字本身,而是它把\"用户—硬件—模型供给\"三层结构串起来。当用户规模过 7 亿、智能硬件渗透率近 40%、国产开源模型占据全球调用前列这三件事同时发生,中国 AI 产业链的\"软硬一体\"逻辑就从 PPT 走进了 CNNIC 数据。","https:\u002F\u002Fwww.cnnic.cn\u002Fn4\u002F2026\u002F0930\u002Fc326-11690.html","4f9c537a-c7bd-44b3-a91e-c0fd224f5cc5",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",{"id":19,"name":20,"slug":20,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":22,"name":23,"slug":23,"description":14,"color":14},"8ddf2b28-0234-41a4-9862-3f0faef96472","market-analysis",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"528d6edf-cbf9-4b9b-a77e-e9b8ee3612e2","en","CNNIC: China's generative AI users topped 700M in H1 2026, penetration over 50%","CNNIC's Generative AI Applications Development Report (2026) shows China's generative AI user base exceeded 700 million by H1 2026, with penetration over 50%. 76% of users rely on AI for Q&A, 47.8% for image and video work, and 38.7% of netizens bought smart hardware in the past six months. The report says the top six models on the global mainstream LLM call rankings all come from Chinese teams.","## Report data: how big did China's generative AI grow in six months\n\nCNNIC's Generative AI Applications Development Report (2026) gives a panoramic figure: by H1 2026, China's generative AI user base broke through 700 million, with penetration above 50%. That means roughly one in every two Chinese netizens now uses a generative AI tool, while half a year ago the figure was still under the halfway mark. CNNIC draws this curve pretty steep.\n\nBreaking down user behavior, 76.0% of users say they let generative AI answer questions — Q&A is still the dominant entry point. The real incremental growth comes from multimodality: 47.8% use AI for image and video work, 37.6% for text generation, and 32.5% write AI into work summaries, meeting notes, and PPT workflows. AI has seeped into the three high-frequency office scenes.\n\n## Hardware entry: wearables and 3C digital gear are the penetration battlefield\n\nAnother frequently overlooked figure in the report — 38.7% of netizens bought smart hardware devices online in the past six months. Wearables account for 20.2%, while smartphones, tablets, and other 3C digital products take 18.2%. The soft-hard collaboration path for generative AI is not starting from new form factors like AR glasses, but eating its first wave of dividends from already mature smart watches and mobile devices.\n\nLining these two figures up, the logic becomes clear: wearables and 3C digital gear are the hardware entry, generative AI is the software entry, both ends accelerating. A user may have bought a new smart watch and frequently call the AI assistant from their phone — both ends got official endorsement in the CNNIC report.\n\n## Chinese vendors take the entire top six on the call rankings\n\nThe report's most sensitive data point: on the global mainstream LLM call rankings, the top six models all come from Chinese teams. The report names DeepSeek and Moonshot (kimi), noting that local firms have successively released multiple trillion-parameter open-source large models.\n\nThis claim requires careful reading. The specific institution and ranking behind \"call rankings\" is not named in the CNNIC reposted version. But if the data is real, the direction is clear: in the past two years, Chinese large models have firmly stood at the global forefront across engineering metrics like inference call volume, Token consumption, and developer adoption. Open-weight models play a role here that's completely different from closed-source APIs — they provide a low-cost, locally deployable, fine-tunable base that pushes call volumes to the first tier.\n\n## A few judgments beyond the data\n\nFirst, the 700 million users and 50% penetration figures come from CNNIC sample surveys, not operator or payment platform level hard data. The strength is broad coverage; the weakness is the loose definition of frequency — \"have used\" and \"use daily\" are two different things. A more accurate reading of \"broke through 700 million\" is that the number of Chinese netizens who have used generative AI exceeds 700 million.\n\nSecond, 76% Q&A, 47.8% multimodal, 37.6% text, 32.5% office — these ratios can stack on the same user, with the total far exceeding 100%. It shows AI tools are not single-point penetration but scene stacking.\n\nThird, the top-six-all-from-China claim is directly tied to the open-source ecosystem and price war. Open-weight models like DeepSeek-V3, Kimi K2, and the Qwen3 series have pushed inference unit prices down to a fraction of closed-source API costs, with developers voting with their feet — but whether \"global call top six\" counts internal testing traffic, the report does not address.\n\n## So what\n\nWhat makes this CNNIC report most worth reading is not the 700 million figure itself, but how it strings together the three-layer structure of \"users, hardware, model supply.\" When user scale passes 700 million, smart hardware penetration approaches 40%, and domestic open-source models occupy the global call rankings' top spots, all happening simultaneously, the soft-hard-integration logic of China's AI industry chain moves from PPT into CNNIC data.","cnnic-china-genai-users-700m-2026","2026-10-14T04:00:00Z","2026-10-14T09:07:54.447402Z","2026-10-14T09:07:54.447415Z",true,"agent",209,[39,48],{"slug":40,"tag_slug":40,"title_zh":41,"title_en":42,"intro_zh":43,"intro_en":44,"id":45,"is_active":35,"created_at":46,"modified_at":47},"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":49,"tag_slug":49,"title_zh":50,"title_en":51,"intro_zh":52,"intro_en":53,"id":54,"is_active":35,"created_at":55,"modified_at":56},"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":58},[59,64,69,74,79,84],{"id":60,"title":61,"news_slug":62,"published_at":63},"7d3de6b6-1faf-47fc-a513-62ee8e731aad","CNNIC 2026 报告:7 亿用户、2185 EFLOPS 算力、调用前六全是中国","cnnic-2026-china-generative-ai-700m-users-2185-eflops","2026-09-29T06:00:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"c4375463-f274-464e-998e-6f2a9f9cfeee","Mozilla:中美开放权重AI差距缩至4.4个月","mozilla-china-open-weight-ai-gap-4-4-months","2026-09-26T00:00:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"b7668f43-05a7-46d7-845f-27e70fcaceec","Mozilla 报告:中国开放权重距美国前沿模型只差 4.4 个月","mozilla-open-weights-4-month-gap","2026-09-16T13:08:54+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"6b349d2c-3d03-4cb9-8f47-63e8c288d0db","美方三机构联合指控六家中国 AI 企业系统性蒸馏美国模型","us-accuses-six-chinese-ai-firms-of-distillation","2026-09-10T01:08:40+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"1464179a-2b7f-4369-b680-25868ddd9042","皮尤实测：超过三分之一 ChatGPT 后的英文网页已有 AI 写作痕迹","pew-research-ai-web-content-2026","2026-08-31T03:00:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"5a6c3aea-9376-4a3d-a79f-a0d19406e299","月之暗面 Kimi K3 想从微软、AWS、Google 手里分一杯羹：中美大模型的收益分成时代","moonshot-kimi-k3-cloud-revenue-share-talks","2026-08-28T12:30:00+00:00"]