[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-nvidia-china-developer-day-2026-suzhou":3,"topics-all":38,"news-related-367476a4-b9af-46f1-a6ab-3de1d83640ff":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},"367476a4-b9af-46f1-a6ab-3de1d83640ff","NVIDIA 把中国开发者日搬到苏州:10 月连开两天,AI 推理和物理 AI 是主菜","NVIDIA 宣布 2026 年 10 月 15—16 日在苏州举办中国开发者日,首日动手实践与黑客松总决赛,次日设 LLM 训练与推理、开源模型、物理 AI\u002F机器人、AI 基础设施四个分论坛,并开 4 门 Associate 认证现场考试。","苏州被 NVIDIA 选为 2026 中国开发者日的落地城市,这件事本身就是信号。\n\n10 月 15—16 日两天日程,15 日动手实践,16 日主论坛加分论坛。地点放在苏州金鸡湖国际会议中心 A 馆,呼应的是长三角近两年在 AI 推理基础设施、机器人、具身智能上的密集布局——苏州周边已聚起一批 GPU 服务器、机器人和自动驾驶的硬科技团队。\n\n## 16 日四个分论坛,等于 NVIDIA 给中国 AI 的\"年度技术地图\"\n\n官方把 16 日下午拆成四条平行跑道:**大语言模型训练与推理、开源模型与运行时、物理 AI\u002F机器人、AI 基础设施**。\n\n- **大语言模型训练与推理**专场关键词是\"测试时计算\"——大模型推理从一次性生成走向更深度的多轮推理,如何设计、部署和优化大规模推理系统,是国内做大模型的公司今年集中补课的领域。\n- **开源模型与运行时**专场延续 NVIDIA 推 NeMo、TensorRT-LLM 的路径——卖卡之外,卖\"怎么让开源模型跑得更快\"的工具链。\n- **物理 AI\u002F机器人**专场配合 Cosmos 与世界模型主旨演讲——NVIDIA 把 Cosmos 3 摆在 16 日上午 9:40 的开幕主题位,意图很明确:大语言模型之后,下一波算力消耗的主战场是机器人和具身智能。\n- **AI 基础设施**专场对应 4 门 Associate 认证里的\"AI 基础架构与运维\",面向数据中心和集群运维工程师。\n\n四条线加在一起,差不多就是 NVIDIA 给中国 AI 产业画的\"下一阶段重点投入地图\"。\n\n## 培训日比主论坛更值得看\n\n15 日全天动手实践,4 场培训和实训营,席位有限、报名确认制:\n\n- 9:30—18:00 **NVIDIA Isaac 端到端机器人开发工作流**(全天),讲 sim-to-real、SIL\u002FHIL 在环仿真到真实部署,是机器人工程师最缺的实操路径。\n- 9:30—18:00 **大规模部署与优化 AI 推理**(全天),需 LLM 推理基础 + 容器\u002FK8s 经验。\n- 9:30—12:30 **使用 OpenShell 和 NemoClaw 构建安全智能体**(上午实训营),讲智能体循环、工具调用、多智能体协作和 RAG。\n- 14:00—17:30 **从感知到预测:使用 Cosmos 3 开发物理 AI**(下午实训营),围绕世界推理、合成数据生成和动作生成。\n\nIsaac 和 Cosmos 两条线同时铺开,说明 NVIDIA 押注的不只是\"卖卡\",而是\"卖从仿真到部署的整条工作流\"。\n\n## 现场认证与黑客松\n\n15 和 16 日两天都开 Associate 认证现场考试,4 门可选:生成式 AI(NCA-GENL,60 分钟 50-60 道选择题)、生成式 AI 多模态(NCA-GENM)、加速数据科学(NCA-ADS)、AI 基础架构与运维(NCA-AIIO),考试可选中文或英文。本质是用免费考试换\"开发者被绑定到 NVIDIA 工具链\"的隐性锁定。\n\n15 日 13:30—17:00 是 2026 NVIDIA 黑客松各分赛道优胜团队的总决赛路演,16 日上午颁奖。看决赛路演比 keynote 信息密度高得多。\n\n## 报名门槛\n\n注册走 NVIDIA 官方账户流程,培训限选 1 场,认证限选 1 门 Associate。2025 年名额开放后短时间内就满,建议尽早申请。\n\n苏州这站,既是给长三角 AI 生态的见面礼,也是 NVIDIA 自己在 2026 下半年的技术宣言。","https:\u002F\u002Fwww.nvidia.cn\u002Fdeveloper-day\u002F?ncid=partn-270555","474eef8c-e0c3-46cf-adee-c089558220f9",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"fca9258a-9430-455a-b95d-b9fae5e373a8","ai-inference",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},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":22,"name":23,"slug":23,"description":14,"color":14},"8dac812d-3839-4abe-a855-5f56ec9515fd","nvidia",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"f67b679b-1042-4280-bbd6-cb20435aa972","en","NVIDIA China Developer Day 2026 Heads to Suzhou","NVIDIA China Developer Day 2026, Suzhou Oct 15-16. Day 1: hands-on training plus Hackathon final. Day 2: main forum plus 4 parallel tracks.","NVIDIA picking Suzhou as the venue for its 2026 China Developer Day is itself a signal.\n\nThe two-day agenda on October 15-16 puts hands-on training on Day 1 and a main forum plus parallel tracks on Day 2. The location, Suzhou Jinji Lake International Conference Center Hall A, tracks the Yangtze River Delta dense build-out of AI inference infrastructure, robotics, and embodied intelligence over the past two years. Suzhou already hosts a cluster of GPU server makers, robotics chassis builders, and autonomous-driving hard-tech teams.\n\n## Four Day-2 Tracks: NVIDIA Annual Tech Map for China AI\n\nThe official schedule splits Day 2 afternoon into four parallel tracks: LLM training and inference, open-source models and runtimes, physical AI \u002F robotics, and AI infrastructure.\n\n- The LLM training and inference track centers on test-time compute, moving from one-shot generation toward deeper multi-turn inference, and how to design, deploy, and optimize large-scale inference systems. This is where Chinese model companies are intensely upskilling this year.\n- The open-source models and runtimes track continues NVIDIA NeMo and TensorRT-LLM push, beyond selling GPUs, selling the toolchain that makes open models run faster.\n- The physical AI \u002F robotics track pairs with the Cosmos and world-models keynote. NVIDIA has placed Cosmos 3 in the 9:40 a.m. opening slot on Day 2, signaling clearly: after LLMs, the next compute battleground is robotics and embodied intelligence.\n- The AI infrastructure track lines up with the AI Infrastructure and Operations Associate certification, targeting data-center and cluster operations engineers.\n\nTogether, the four tracks are essentially NVIDIA next-phase investment map for China AI industry.\n\n## Training Day Is More Worth Watching Than the Main Forum\n\nDay 1 is all hands-on. Four training sessions and workshops, with limited seats and registration-by-confirmation:\n\n- 9:30 to 18:00 NVIDIA Isaac End-to-End Robotics Development Workflow (full day), covering sim-to-real, SIL\u002FHIL in-the-loop simulation, and real-world deployment. The most missing hands-on path for robotics engineers today.\n- 9:30 to 18:00 Large-Scale Deployment and Optimization of AI Inference (full day), requires LLM inference fundamentals plus container\u002FKubernetes experience.\n- 9:30 to 12:30 Building Secure Agents with OpenShell and NemoClaw (morning workshop), NVIDIA own framework for agent loops, tool calls, multi-agent collaboration, and RAG.\n- 14:00 to 17:30 From Perception to Prediction: Physical AI with Cosmos 3 (afternoon workshop), world reasoning, synthetic data generation, and action generation.\n\nIsaac and Cosmos running in parallel signals that NVIDIA is betting not just on selling GPUs but on selling the full workflow from simulation to deployment.\n\n## On-Site Certifications and Hackathon\n\nBoth Day 1 and Day 2 run Associate-level certification exams on-site, four options: Generative AI (NCA-GENL, 60 min, 50-60 multiple-choice questions), Generative AI Multimodal (NCA-GENM), Accelerated Data Science (NCA-ADS), and AI Infrastructure and Operations (NCA-AIIO). The exam is offered in both Chinese and English. In essence, this is a free exam in exchange for the hidden lock-in of binding developers to NVIDIA toolchain.\n\nDay 1 13:30 to 17:00 is the 2026 NVIDIA Hackathon grand-final demo for winners from each regional track, with the awards on the morning of Day 2. Watching the final demos carries higher information density than the keynotes.\n\n## Registration Threshold\n\nRegistration flows through the NVIDIA official account. Training is limited to 1 session per person; certification is limited to 1 Associate per person. The 2025 Developer Day filled up fast once registration opened, so apply early.\n\nThe Suzhou stop is both a hello gift to the Yangtze River Delta AI ecosystem and NVIDIA own 2026 second-half technology manifesto.","nvidia-china-developer-day-2026-suzhou","2026-09-16T03:00:00Z","2026-09-16T05:03:29.985649Z","2026-09-16T05:03:29.985665Z",true,"agent",27,[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},"1bf3f192-68e9-49f1-8bbb-a350f05c15e2","LLM 编排云边端:FRAME 2026 论文点出反馈层缺环","llm-cloud-continuum-frame-2609","2026-09-16T11:00:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"61cf8d85-c751-4da2-9aae-10b645415ec9","英伟达发布开源工具 PAIR,把家里电脑连成个人 AI 推理集群","nvidia-pair-personal-ai-router-local-inference","2026-09-09T02:00:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"42b7939c-1b44-43b8-95cf-a8fc2204560d","NVIDIA 开源 Personal AI Router，把家里 RTX 与 Mac 拼成本地 AI 集群","nvidia-personal-ai-router-pair-beta","2026-09-04T03:20:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"64b79fb5-b416-4b58-8402-0bc9ad22a27d","腾讯混元 Hy ASR 3.0 preview：MoE + 语音 Encoder 与 LLM 联合训练，把 ASR 推进到'理解语境'","tencent-hunyuan-hy-asr-3-0-preview-moe-speech-lm","2026-08-04T12:00:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"30a147aa-e3ed-475c-b15f-9e5ffce6ffc9","英伟达 Vera CPU：DeepInfra 实测 Agent 编排提速 2.2 倍","nvidia-vera-cpu-agent-orchestration","2026-07-22T04:50:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"f4aad332-1f97-4cb2-96a4-37d8d2980728","英伟达BW首秀RTX Spark：笔记本本地跑120B大模型","nvidia-rtx-spark-120b-laptop","2026-07-12T12:00:00+00:00"]