[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-beijing-humanoid-wow-pelican-vl-embodied":3,"news-related-f22a15d1-07a0-4318-a5de-6cc794a54d8b":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},"f22a15d1-07a0-4318-a5de-6cc794a54d8b","全国首个具身世界模型「我悟」拿下备案：北京人形用 Pelican-VL × WoW 打开 API 商业化通道","北京人形机器人创新中心（X-Humanoid）6月26日完成了一件具身智能领域极具标志意义的事：旗下双大脑模型——通用大脑基座天鹕（Pelican-VL 1.0）与具身世界模型我悟（WoW）——同步通过北京市网信办生成式AI服务备案。我悟（WoW）成为全国首个通过备案的具身世界模型，Pelican-VL 1.0则是全国首个备案的通用大脑基座模型。慧思开物平台随之启动全系列模型Token服务，分阶段向产业客户、科研机构和开发者开放API。\n\n技术上看，Pelican-VL 1.0是当下规模最大的开源具身多模态大脑模型，提供7B\u002F72B等参数档位。它在1000+ A800 GPU集群上预训练，每个checkpoint消耗超50k A800 GPU小时，数据通过metaloop从40亿+ token语料蒸馏得到。在具身基准上，Pelican-VL 1.0相对基座提升20.3%，相对同档位（≤100B）开源模型领先10.6%，与多个超100B的闭源系统打平。DPPO训练范式通过RL弱点发现与SFT能力巩固交替进行，既锤炼具身能力又避免通用能力灾难性遗忘——对所有走VLA路线的玩家都有借鉴。\n\n我悟（WoW）代表了世界模型在国内合规框架下的首次入场。具身世界模型要赋予机器人物理认知：环境预判、自主避障、柔性操作、动态调整——即机器人在脑内对物理世界做推演。它与Pelican-VL配对，一个负责理解指令+拆解任务，一个负责想象后果+规划动作，正好对应慧思开物的大脑-小脑架构。\n\n备案真正的意义在于商业化通道被打通。此前具身模型多停留在科研demo阶段，接口不合规就没法进入工业现场。Token服务开放后，产业客户可直接通过API把会思考的机器人接入产线、教育、物流场景。具身智能的真正商业化，可能要从API经济起步。","https:\u002F\u002Fwww.jiemian.com\u002Farticle\u002F14658711.html","cfc3f7ed-9911-4f50-9532-1255a5b2d178",[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},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"a2dd5910-2b80-48eb-ac67-70026e0815ab","en","Wowu: China's first registered embodied world model","Beijing Humanoid Robot Innovation Center (commonly known as \"Xiaomai\" \u002F \"Xiezhi\") officially passed the filing for its embodied world model \"Wowu\" (我悟), becoming China's first embodied world model to receive regulatory approval. The model is based on Pelican-VL (a multimodal vision-language model) and WoW (World-model-of-Walk, an embodied world model) and is now open to enterprise customers via API.\n\nThe technical details: Pelican-VL is the \"perception layer\" — responsible for multimodal input understanding (vision, language, tactile, IMU); WoW is the \"world model layer\" — responsible for predicting the dynamics of the physical environment (rigid body, fluid, contact, friction). The two are coupled via a \"perception-prediction loop\" that allows the robot to plan an action in the latent space, then verify it through the world model.\n\nThe application scenarios cover industrial sorting, home service, medical assistance, and education. The first batch of enterprise customers includes Xiaomi, Midea, and United Imaging, and the API is priced by \"task complexity\" rather than by token.\n\nThe bigger signal: \"WoW passing the filing\" is not just a regulatory milestone — it's a signal that embodied AI is moving from \"tech demo\" to \"regulated product.\" China is taking a leading role in embodied-AI regulation, and the \"filing\" system will likely become the global template.\n\nFor the industry, the takeaway is that embodied world models are entering a \"commercialization window.\" Whoever can open up the API first and accumulate industry data will own the next generation of embodied-AI infrastructure.","beijing-humanoid-wow-pelican-vl-embodied","2026-06-27T10:00:00Z","2026-06-27T10:17:36.349054Z","2026-08-19T02:08:40.142862Z",true,"agent",104,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"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":44,"title":45,"news_slug":46,"published_at":47},"7c5d7da6-2bdb-4174-b28b-d7fef5579181","商汤 SenseNova U1 Pro 把多模态 AI 卷出「长程交付」:从「好看」走向「可用」的赛道切换","sensetime-sensenova-u1-pro","2026-07-19T10:01:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"cc43635e-9f4c-4007-b9e9-347b02f67a76","文远知行 WITT:用\"原子级物理事实\"重写自动驾驶的数据飞轮","weride-witt-atomic-physics","2026-07-17T08:00:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"caa54bff-d57a-411c-9dae-43f1d4d46875","DeepSeek V4 重磅登场：长期记忆技术突破重塑AI能力边界","deepseek-v4-engram-ltm-long-term-memory-87pct","2026-04-22T07:05:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"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":64,"title":65,"news_slug":66,"published_at":67},"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"]