[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-cnnic-2026-china-generative-ai-700m-users-2185-eflops":3,"topics-all":44,"news-related-7d3de6b6-1faf-47fc-a513-62ee8e731aad":63},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":30,"news_slug":37,"published_at":38,"created_at":39,"modified_at":40,"is_published":41,"publish_type":42,"image_url":14,"view_count":43},"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 9 月 29 日发布《生成式人工智能应用发展报告(2026)》,2026 上半年我国生成式 AI 用户突破 7 亿、普及率超 50%。报告同时披露智能算力 2185 EFLOPS、同比增长 177%,全国产 10 万卡 AI 超集群投用,中国团队包揽全球大模型调用榜前六。","中国互联网络信息中心(CNNIC)9 月 29 日在北京发布的《生成式人工智能应用发展报告(2026)》把上半年的中国 AI 行业一并摊在桌面上:用户端、算力端、模型端、产业端、资本端五条线同时给出新数字,既是半年总结,也是下半年起跑的基线。\n\n## 用户端:半数网民已经在用\n\n报告测算截至 2026 年上半年,我国生成式 AI 用户规模突破 7 亿、普及率超过 50%。76% 的用户表示自己会让生成式 AI 回答问题,处理图片\u002F视频、文本和工作总结\u002F会议纪要\u002FPPT 的占比分别为 47.8%、37.6% 和 32.5%。AI 综合助手、AI 效率办公等应用的使用次数同比增长均超过 100%。\n\n硬件侧同样在加速渗透。38.7% 的网民近半年在网上购买过智能硬件设备,其中可穿戴设备和 3C 数码产品是首要入口,购买智能可穿戴设备的比例为 20.2%,购买智能手机、平板电脑等 3C 数码产品的比例为 18.2%。这意味着 AI 已经从屏幕里的对话框延伸到手腕、耳朵和口袋里。\n\n## 算力端:2185 EFLOPS 与全国产 10 万卡\n\n截至 2026 年 6 月,我国智能算力规模达到 2185 EFLOPS,同比增长 177%。同月,首个全国产 10 万卡人工智能超集群正式投用,在芯片、计算、存储、网络、散热等核心软硬件环节全部采用国产自主可控技术,实现从底层硬件到上层服务的全栈国产化。这是国内首个进入\"十万卡级\"部署的 AI 算力底座。配合 12 万个以上已建成的高质量数据集,模型训练所需的\"算力 + 数据\"基本盘在国内首次形成闭环。\n\n## 模型端:中国团队包揽调用榜前六\n\n报告给出一个外界关切的明确信号: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。DeepSeek、千问等开源模型的影响力在全球范围内持续扩张,Qwen 3.8 Max 已经以 2.4 万亿参数在 9 月开源,加上 Qwen 3.8-Flash-Next 把激活参数压到 6B、训练成本压到上一代的 1\u002F9,中国阵营在\"超大模型\"和\"高效率模型\"两条路线上同时握有主力选手。\n\n## 产业与资本端:半年融资已是去年 1.8 倍\n\n报告披露我国已累计培育 16 个国家级先进制造业集群,规上制造业企业人工智能技术应用普及率已超过 30%,AI 贯穿研发设计、生产制造、质检运维等全链条。资本端,2026 年上半年我国发生 1255 起与人工智能相关的投融资事件,合计金额约 2501.4 亿元,半年融资额已经达到 2025 年全年的 182%。供给端,截至 2026 年 6 月,我国累计有 988 款生成式人工智能服务在国家网信办完成备案,598 款在地方网信办完成登记。\n\n把这五条线拼到一起看,报告实际回答的是一个老问题:中国 AI 的\"过半拐点\"到了没有?用户侧 50% 普及率、算力侧 2185 EFLOPS 和 10 万卡国产集群、产业侧 30% 规上制造业渗透率、资本侧半年融资超 2500 亿——按创新扩散的一般规律,普及率接近 50% 意味着开始进入增长拐点,新用户增长会放缓,行业重心从规模扩张转向质量提升和价值创造。这场\"AI 用户破 7 亿\"的故事,真正的下半场不在拉新,在落地。","https:\u002F\u002Fwww.news.cn\u002Fpolitics\u002F20260929\u002F2b4afc8c85b14ff09c50a6a67ce2bf29\u002Fc.html","46f25792-35bc-455d-bd12-9f30ad99d2e6",[11,15,18,21,24,27],{"id":12,"name":13,"slug":13,"description":14,"color":14},"c33b1bbc-d6ce-4f61-9d5d-1a0704a6a09b","ai-policy",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",{"id":19,"name":20,"slug":20,"description":14,"color":14},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",{"id":22,"name":23,"slug":23,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":25,"name":26,"slug":26,"description":14,"color":14},"8ddf2b28-0234-41a4-9862-3f0faef96472","market-analysis",{"id":28,"name":29,"slug":29,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[31],{"id":32,"lang":33,"title":34,"summary":35,"content":36},"a959066a-d5aa-4718-8d0f-fdcfec2e744a","en","CNNIC: 700M Chinese AI Users, Top-6 LLM Calls Go to China","CNNIC: H1 China AI users hit 700M (50%+); intelligent compute hit 2,185 EFLOPS (+177% YoY); Chinese teams own all top-6 global LLM call rankings.","China's Internet Network Information Center (CNNIC) released the Generative AI Application Development Report (2026) at the 7th China Internet Infrastructure Resources Conference on September 29 in Beijing. The report draws a half-year baseline across five lines—users, compute, models, industry, and capital—that will define the second half of 2026.\n\n## Users: half of all netizens are already on board\n\nAs of H1 2026, China's generative AI user base crossed 700 million, with penetration above 50%. 76% of users say they let generative AI answer questions for them, while 47.8%, 37.6%, and 32.5% use it for image\u002Fvideo processing, text work, and meeting notes\u002Fslide drafts respectively. Monthly usage of AI general assistants and AI efficiency-office apps grew by more than 100% year-on-year.\n\nOn the hardware side, 38.7% of netizens bought a smart hardware device in the past six months—wearables and 3C electronics are the leading entry points, at 20.2% and 18.2% respectively. Generative AI is no longer just a chat window; it is moving onto wrists, ears, and into pockets.\n\n## Compute: 2,185 EFLOPS and a fully-domestic 100,000-card super-cluster\n\nAs of June 2026, China's intelligent compute capacity reached 2,185 EFLOPS, up 177% year-on-year. In the same month, the country's first fully-domestic 100,000-card AI super-cluster went live, with chips, compute, storage, networking, and cooling all built on domestic self-controllable technology, marking end-to-end localization from silicon to service. This is the first domestic AI compute base to cross into the \"hundred-thousand-card\" tier. Combined with more than 120,000 high-quality datasets already in place, the \"compute + data\" loop required for training frontier models is closing inside China for the first time.\n\n## Models: Chinese teams own all top-6 call rankings\n\nThe report delivers an explicit signal the industry has been waiting for: all top-6 models on the global mainstream LLM call rankings are now from Chinese teams. Open-source models led by DeepSeek and Qwen continue to expand globally: Qwen 3.8 Max released in September at 2.4 trillion parameters, alongside Qwen 3.8-Flash-Next with 6B activated parameters and only one-ninth of the previous generation's training cost. Chinese players now hold flagship cards on both the \"ultra-large\" and \"high-efficiency\" tracks.\n\n## Industry and capital: H1 funding already 1.82x of last year\n\nThe report discloses 16 national-level advanced manufacturing clusters have been cultivated, AI penetration among regulated manufacturing firms has crossed 30%, and AI is woven through R&D, production, QA, and operations. On the capital side, 1,255 AI-related investment events occurred in H1 2026, totaling roughly 250.14 billion yuan—already 182% of the full-year 2025 total. On the supply side, 988 generative AI services have registered with the Cyberspace Administration of China, with another 598 applications registered at the provincial level by June 2026.\n\nStack these five lines and the report answers one question: has China's AI hit the \"50% tipping point\"? With 50% user penetration on the demand side, 2,185 EFLOPS plus a 100,000-card domestic cluster on the supply side, 30% manufacturing AI penetration, and H1 funding past 250 billion yuan, the answer is yes—diffusion-of-innovation theory says crossing 50% means a turning point where new-user growth slows and the center of gravity shifts from scale to quality and value creation. The second half of the \"AI users cross 700M\" story will be measured by what gets grounded in production, not by another round of onboarding.","cnnic-2026-china-generative-ai-700m-users-2185-eflops","2026-09-29T06:00:00Z","2026-10-01T03:07:29.341037Z","2026-10-01T03:07:29.341048Z",true,"agent",199,[45,54],{"slug":46,"tag_slug":46,"title_zh":47,"title_en":48,"intro_zh":49,"intro_en":50,"id":51,"is_active":41,"created_at":52,"modified_at":53},"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":55,"tag_slug":55,"title_zh":56,"title_en":57,"intro_zh":58,"intro_en":59,"id":60,"is_active":41,"created_at":61,"modified_at":62},"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":64},[65,70,75,80,85,90],{"id":66,"title":67,"news_slug":68,"published_at":69},"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":71,"title":72,"news_slug":73,"published_at":74},"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":76,"title":77,"news_slug":78,"published_at":79},"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":81,"title":82,"news_slug":83,"published_at":84},"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",{"id":86,"title":87,"news_slug":88,"published_at":89},"1844afb1-3a1c-4acd-9e4c-f5e2792a2018","下载免费不等于商用免费：HF Summer 2026 隐藏的开源前沿许可证分水岭","frontier-license-shift-hf-summer-2026","2026-08-23T12:30:00+00:00",{"id":91,"title":92,"news_slug":93,"published_at":94},"8d7b30e0-996f-4141-8501-8f464bda6282","中美开放权重参数上限差距拉到 20 倍:Hugging Face 夏季报告里的三条隐藏数据","hugging-face-summer-2026-frontier-ceiling","2026-08-22T14:00:00+00:00"]