广告
加载中

从实验室到万亿市场:智谱与银河通用的“双龙头”进化论

白华 2026-07-02 08:47
白华 2026/07/02 08:47

邦小白快读

EN
全文速览

本文核心介绍北京AI产业形成智谱(数字大模型)+银河通用(具身智能机器人)的双龙头格局,二者共同冲击万亿级通用人工智能市场,核心干货信息整理如下:

1.两家企业核心情况:智谱创始团队来自清华,已经在港交所上市,市值突破万亿港元,其推出的GLM-5.2旗舰大模型主攻长程任务,在全球盲测中位列第一,API涨价后调用量反而暴涨,商业化验证成功。银河通用创始团队来自北大,聚焦具身智能,研发了全球首个端到端具身大模型银河星脑,拿下多个全球首创技术突破,已经在工业、零售、医疗实现落地,累计工业订单达数千台。

2.行业整体现状:北京海淀已经集聚两千多家AI相关企业、三百三十多家具身智能企业,形成了完整全产业链,中国AI已经从单点竞争走向系统竞争,未来数字+物理双轨并行将成为主流方向。

本文披露了国内AGI产业的最新格局与发展趋势,能为品牌商的产品研发、市场布局提供多维度参考,核心干货整理如下:

1.产业与消费趋势:当前通用人工智能已经形成数字大模型+具身智能双路径协同的产业闭环,AI正在从数字领域快速渗透到物理消费场景,零售、医疗、工业制造等多领域都已经有成熟落地案例,给品牌跨界布局指明了新方向。

2.产品研发与定价参考:智谱的商业化数据显示,拥有核心技术壁垒的产品,即使API提价83%,调用量反而增长400%,说明核心技术能支撑价格上涨,无需靠低价换市场;银河通用不追逐社交媒体demo爆款,坚持做工业级落地产品的研发节奏,也给品牌研发提供了参考。

3.新场景机会:具身智能已经实现线下门店24小时自主运营,给品牌布局无人线下零售场景提供了成熟的新方案。

本文披露了AGI产业最新的政策动向与市场机会,给涉足AI领域的卖家提供了明确的参考方向,核心干货整理如下:

1.政策红利明确:北京明确将人工智能打造为下一个万亿级产业集群,提出要统筹推进基础大模型与具身智能协同攻关,国家大基金三期已经首次投资具身智能企业银河通用,政策端对AI核心技术路线的支持信号清晰。

2.市场增长机会清晰:当前北京已经形成从基础大模型到具身智能的完整全产业链,To B端工业制造、To C端零售医疗都有大量真实落地需求,银河通用仅工业机器人订单就达数千台,零售端已经落地两百多家门店,赛道增长空间充足。

3.经营风险提示与可借鉴策略:不要盲目跟风轻量化快迭代风口,不要只追求demo流量,要像银河通用一样,对每个落地场景做技术可行性、工程可交付性、商业可持续性三重验证,聚焦核心场景打磨技术,避免陷入只有流量没有订单的陷阱。

本文介绍了具身智能在工业生产领域的最新落地进展,给工厂推进智能化升级带来了明确的启示和商业机会,核心干货整理如下:

1.生产升级的新机会:当前具身智能重载机器人已经实现7×24小时在新能源量产线常态化自主作业,能适应复杂非结构化工厂环境,无需二维码、反光板定位就能完成物料转运等长程任务,目前已经进入宁德时代、博世、多家主流车企的真实产线,工厂可引入这类机器人提升生产效率,缓解人力成本压力。

2.智能化升级的路径启示:银河通用的技术打磨路径是先在高难度工业场景验证技术壁垒,再逐步拓展场景,工厂做智能化升级不要盲目跟风铺量,要结合自身生产场景优先验证技术适配性,再逐步落地,避免资源浪费。

3.产业协同启示:当前AI产业已经形成大模型+机器人的协同产业生态,工厂可以依托现有成熟产业资源,快速对接成熟的技术方案,推进自身数字化、智能化转型,降低自研技术的成本风险。

本文梳理了AGI产业的最新发展趋势,明确了产业链上的客户痛点与可行解决方案,给AI相关服务商指明了发展方向,核心干货整理如下:

1.行业发展趋势清晰:AGI已经分化为数字大模型和具身智能两大核心路径,二者将长期协同发展,北京已经形成覆盖从基础层到应用层的完整全产业链,未来AI将成长为万亿级产业,产业链各环节服务商都有充足的市场空间。

2.明确了客户核心痛点:传统机器人采用感知-规划-控制三级分离架构,误差和延迟逐层累积,无法适应真实复杂的非结构化生产运营场景;传统大模型只能完成即时问答,无法支撑连续长程的大型任务,这些都是当前客户的核心痛点。

3.成熟解决方案参考:银河通用推出的端到端银河星脑具身大模型,打通了全链路神经网络,解决了传统机器人的架构痛点,已经验证了多场景落地效果;智谱推出的GLM-5.2主攻长程任务,能支撑完整大型工程连续运行,服务商可围绕这些成熟技术拓展落地服务。

本文披露了北京AI产业的发展格局与企业需求,给AI领域的平台商运营布局提供了参考,核心干货整理如下:

1.明确了产业对平台的核心需求:当前AI已经从单点技术竞争走向产业系统竞争,需要平台统筹基础大模型与具身智能的协同攻关,搭建多层次、多梯队的创新矩阵,满足不同技术路线企业的协作需求,完善软硬件技术底座。

2.平台招商与运营可借鉴海淀AI集群的经验:依托区域丰富的创新资源,引入不同技术路线的龙头企业,分别打造数字大模型和具身智能的龙头,形成龙头引领、中小企业跟进的多层次创新矩阵,构建覆盖全产业链的产业生态,吸引更多相关企业入驻。

3.风险规避方向:平台招商布局要规避只做流量demo、没有真实落地能力的项目,要优先支持深耕核心技术、已经通过商业化落地验证的项目,避免产业泡沫,降低平台运营风险。

本文梳理了中国AGI产业发展的最新动向与商业模式创新,对AI产业研究有很高的参考价值,核心干货整理如下:

1.产业发展新动向:中国AI已经完成从单点竞争到系统竞争的关键跃迁,北京形成了数字大模型+具身智能双龙头的产业格局,海淀已经构建了从基础大模型到具身智能的完整全产业链,全球罕见的数字+物理双轨并行的产业闭环已经形成。

2.技术发展新突破:智谱的GLM-5.2大模型实现了长程任务突破,在全球百万用户盲测中位列全球可用模型第一;银河通用提出WAM技术路线被业内称为机器人技术终局,实现了全自主网球对拉、真实机器人转笔等多个全球首创技术突破,还验证了人形机器人在量产线常态化作业,业内判断具身智能的ChatGPT时刻即将到来。

3.商业模式创新:当前已经形成两种成熟可复制的AGI商业模式,智谱的Token经济模式,按API调用计量智能价值;银河通用的物理经济模式,靠落地机器人提升生产效率创造价值,二者相互协同,为AGI商业化提供了新路径。

返回默认

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

This article introduces the dual-leading-firm pattern that has emerged in Beijing's AI industry: Zhipu AI for digital large language models (LLMs) and Galaxy General Robotics for embodied intelligent robots. Together, the two companies are targeting the trillion-yen general artificial intelligence (AGI) market. Key takeaways are as follows:

1. Core profiles of the two firms: Founded by a team from Tsinghua University, Zhipu AI is listed on the Hong Kong Stock Exchange with a market capitalization exceeding HK$1 trillion. Its flagship GLM-5.2 model is optimized for long-horizon tasks, ranking first in global blind testing. Even after a price hike for its API services, call volume has surged sharply, proving its commercial viability. Galaxy General Robotics, founded by a Peking University-origin team, focuses on embodied AI and has developed the world's first end-to-end embodied large model, Galaxy Brain, with multiple world-first technological breakthroughs. Its solutions are already deployed in industrial, retail, and healthcare sectors, with cumulative industrial orders reaching thousands of units.

2. Current status of the broader industry: Haidian District of Beijing has gathered more than 2,000 AI-related enterprises and over 330 embodied AI firms, forming a complete full industrial chain. China's AI sector has shifted from single-point competition to systemic competition, and the parallel development of digital and physical AI will become the mainstream direction going forward.

This article outlines the latest landscape and development trends of China's AGI industry, providing multi-dimensional insights for brand product R&D and market layout. Key takeaways are as follows:

1. Industrial and consumer trends: The AGI industry has formed a closed loop of dual-path development integrating digital large models and embodied intelligence. AI is rapidly expanding from the digital domain to physical consumer scenarios, with mature deployments already in retail, healthcare, industrial manufacturing and other sectors, pointing out new directions for cross-sector brand expansion.

2. References for product R&D and pricing: Zhipu AI's commercial data shows that for products with core technical barriers, even an 83% API price hike led to a 400% increase in call volume. This proves core technology can support price increases, eliminating the need to compete on low prices. Galaxy General Robotics' strategy of rejecting social media demo hype and sticking to a R&D rhythm focused on industrial-grade deployable products also provides a useful reference for brand R&D.

3. New scenario opportunities: Embodied AI has enabled 24-hour autonomous operation of offline stores, offering a mature new solution for brands looking to layout unmanned offline retail scenarios.

This article discloses the latest policy trends and market opportunities in the AGI industry, providing clear directional references for sellers active in the AI space. Key takeaways are as follows:

1. Clear policy dividends: Beijing has explicitly positioned artificial intelligence as its next trillion-yuan industrial cluster, and is coordinating collaborative R&D for both foundational large models and embodied intelligence. The third tranche of China's National Fund has made its first investment in an embodied AI firm, Galaxy General Robotics, sending a clear signal of policy support for this core AI technology route.

2. Clear market growth opportunities: Beijing has built a complete full industrial chain ranging from foundational large models to embodied intelligence, with strong real-world demand across B2B industrial manufacturing and B2C retail and healthcare. Galaxy General alone has thousands of industrial robot orders and deployments in more than 200 retail stores, leaving substantial room for growth in the sector.

3. Risk warnings and replicable strategies: Sellers should not blindly follow the trend of lightweight, fast-iteration projects or chase only demo-driven traffic. Instead, they should follow Galaxy General's example: conduct triple verification of technical feasibility, engineering deliverability and commercial sustainability for every deployment scenario, focus on core scenarios to refine technology, and avoid the trap of having only traffic but no substantive orders.

This article introduces the latest deployment progress of embodied intelligence in industrial production, bringing clear insights and commercial opportunities for factories advancing intelligent upgrading. Key takeaways are as follows:

1. New opportunities for production upgrading: Heavy-duty embodied intelligent robots now achieve 7/24 autonomous routine operation on new energy mass production lines. They can adapt to complex unstructured factory environments and complete long-horizon tasks such as material transfer without relying on QR codes or reflector positioning. These robots are already deployed on the actual production lines of CATL, Bosch, and multiple leading automakers. Factories can introduce such robots to improve production efficiency and ease labor cost pressure.

2. Insights for intelligent upgrading paths: Galaxy General refines its technology by first verifying technical barriers in high-difficulty industrial scenarios, then gradually expanding to other scenarios. When carrying out intelligent upgrading, factories should not blindly rush to scale; instead, they should prioritize verifying technical adaptability for their own production scenarios before gradual deployment to avoid wasting resources.

3. Insights for industrial collaboration: The AI industry has now formed a collaborative ecosystem integrating large models and robots. Factories can leverage existing mature industrial resources to quickly access proven technical solutions to advance their digital and intelligent transformation, reducing the cost and risk of independent R&D.

This article sorts out the latest development trends of the AGI industry, clarifies core customer pain points along the industrial chain and proven feasible solutions, pointing out development directions for AI-related service providers. Key takeaways are as follows:

1. Clear industry development trends: AGI has split into two core development paths: digital large models and embodied intelligence, which will co-develop and synergize over the long term. Beijing has already built a complete full industrial chain covering layers from the foundation to application. AI will grow into a trillion-yuan industry, leaving sufficient market space for service providers at every link of the industrial chain.

2. Clarified core customer pain points: Traditional robots use a three-stage separated architecture of perception-planning-control, where errors and delays accumulate layer by layer, making them unable to adapt to complex, unstructured real-world production and operation scenarios. Traditional large models can only handle on-the-spot question answering, and cannot support continuous long-horizon large-scale tasks. These are the core pain points facing customers today.

3. References for mature solutions: Galaxy General's end-to-end embodied large model Galaxy Brain connects the entire neural network pipeline, solving the architectural pain point of traditional robots, with proven deployment results across multiple scenarios. Zhipu's GLM-5.2 is optimized for long-horizon tasks and can support the continuous operation of large-scale complete engineering projects. Service providers can expand deployment services around these proven mature technologies.

This article discloses the development landscape and enterprise demand of Beijing's AI industry, providing references for operational layout of AI-focused platform operators. Key takeaways are as follows:

1. Clarified core industrial demand for platforms: AI has shifted from single-point technological competition to industrial systemic competition. Platforms are needed to coordinate collaborative R&D between foundational large models and embodied intelligence, build a multi-level, multi-echelon innovation matrix, meet collaboration needs for enterprises following different technology routes, and improve the software and hardware technology base.

2. Insights for investment attraction and operation from Haidian AI cluster: Platforms can leverage local rich innovation resources to introduce leading enterprises following different technology routes, cultivate separate leading players for digital large models and embodied intelligence, form a multi-level innovation matrix led by top players with small and medium-sized enterprises following, build a full-industry-chain ecological system, and attract more relevant enterprises to settle in.

3. Directions for risk mitigation: When attracting projects for layout, platforms should avoid projects that only focus on traffic-generating demos without real deployment capabilities, and prioritize supporting projects that deepen core technology and have already been validated through commercial deployment, to avoid industrial bubbles and reduce platform operational risk.

This article sorts out the latest developments and business model innovations of China's AGI industry, offering high reference value for AI industry research. Key takeaways are as follows:

1. New trends in industrial development: China's AI industry has completed a key transition from single-point competition to systemic competition. Beijing has formed a dual-leading industrial pattern of digital large models plus embodied intelligence, and Haidian District has built a complete full industrial chain from foundational large models to embodied intelligence. A globally rare closed industrial loop featuring parallel development of digital and physical AI has taken shape.

2. New technological breakthroughs: Zhipu's GLM-5.2 large model achieved a breakthrough in long-horizon tasks, ranking first among all available global models in a blind test with one million global users. Galaxy General's WAM technical route is regarded by the industry as the endgame of robotic technology. It has achieved multiple world-first technological breakthroughs, including fully autonomous tennis rallies and pen-spinning with real robots, and has validated the regular operation of humanoid robots on mass production lines. The industry widely believes that the "ChatGPT moment" for embodied intelligence is coming soon.

3. New business model innovations: Two mature, replicable AGI business models have emerged. Zhipu operates a token-based economic model that measures intelligent value by API calls; Galaxy General operates a physical economic model that creates value by delivering deployed robots to improve production efficiency. The two models complement each other, opening up new paths for AGI commercialization.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

近日,北京市市长殷勇时隔一年再次调研智谱华章与银河通用机器人,这或许是北京AI产业格局的一次“官宣”。一个在数字世界写代码,一个在物理世界造具身身体。两家企业技术路径不同,却共同指向同一个终点:通用人工智能(AGI)。

殷勇在智谱听取了新一代旗舰大模型GLM-5.2的技术汇报,随后又察看了银河通用人形机器人在真实场景中的全自主作业和网球对打互动。

他强调,北京拥有全国最为丰富的人工智能创新资源,正着力将人工智能打造为下一个万亿级产业集群。海淀区目前已集聚2000余家人工智能相关企业和330余家具身智能机器人相关企业,基本形成了从基础大模型到具身智能的全产业链条。

而智谱与银河通用,正是这条产业链上两个最具标志性的“龙头”,一个代表“数字大脑”路径,一个代表“物理身体”路径。它们的平行崛起,揭示了中国AI从单点竞争走向系统竞争的关键跃迁。

01 两种创业方式 同一个AGI终点

2023年5月,北京银河通用机器人在海淀注册成立。创始人兼CTO王鹤,刚满三十岁。他是清华大学本科、斯坦福大学博士,师从美国三院院士Leonidas J.Guibas教授,博士毕业后回国加入北京大学计算机学院前沿计算研究中心,担任研究员、博士生导师。

王鹤的研究方向从一开始就踩在技术前沿上,始终聚焦具身智能与通用机器人模型。在斯坦福期间,他提出的NOCS(标准化物体坐标空间)模型改写了机器人抓取能力的发展轨迹。回国后,他在北大创立了中国第一个以“具身”命名的实验室——具身感知与交互实验室(EPIC Lab),开始了从学术研究到产业化的探索。

在团队构建上,银河通用走了一条少有人走的路,不盲目堆人头,而是以“精兵强将”策略,汇聚了来自清华、北大、北航等顶尖院校的科研骨干,以及微软亚洲研究院等知名机构的行业专家,覆盖具身大模型、仿生结构设计运动控制、计算机视觉等核心领域。

有投资人在接受采访时评价:“这是一支由中国最顶尖90后科研与工程力量组成的队伍,在具身智能的赛道上,他们有资格与世界任何一支团队正面竞争。”

创业之初,银河通用没有选择行业普遍追捧的“轻量化、快迭代”路线,而是选择了啃最硬的骨头。

“我们不追求机器人会做100件事,而是让它把10件事做到工业级标准。”王鹤在多个场合反复强调这一逻辑。这一判断根植于他对技术落地节奏的深刻理解,当整个行业都在追逐社交媒体上的demo爆款时,真正的护城河在于能否让机器人在真实产线上24小时不间断地真干活。

从融资节奏来看,资本市场对这一策略给出了明确的正反馈。三年时间,银河通用成长为中国累计融资额、估值最高的具身智能企业。2026年3月完成的新一轮融资中,由国家人工智能产业投资基金领投,这是国家大基金三期首次投资具身智能企业,被市场解读为“具身大模型国家队”正式确立。

银河通用攻克具身智能核心壁垒,是人工智能深耕物理世界、持续突破的标杆。

另一家企业智谱,成立于2019年,创始团队几乎全部出自清华大学计算机系知识工程实验室(KEG)。创始人兼首席科学家唐杰教授,是国内最早启动通用大模型产业化的学者之一。2026年1月,智谱以116.2港元/股的发行价在港交所上市,半年后股价累计涨幅近20倍,市值一度突破1.07万亿港元,成为国内首家迈入“万亿俱乐部”的大模型企业。

智谱的创业路径是一条典型的“数字大脑”进化路线:从清华大学KEG实验室的技术成果转化起步,到开源GLM系列大模型,再到构建“Token经济—长程任务—企业级应用”的商业闭环。

2026年6月17日,智谱上线并开源新一代旗舰大模型GLM-5.2。该模型主攻“长程任务”,让AI不再只做即时问答,而能像人一样连续工作数小时、自主跑完一个完整的大型工程。在全球百万用户参与盲测的Code Arena榜单上,GLM-5.2位列全球可用模型第一。特斯拉CEO马斯克与唐杰就此在社交媒体上的隔空对话,更是将这场技术叙事推向了全球视野。

财报显示,2025年智谱开放平台及API营收占比从15.5%提升至26.3%,企业级智能体营收占比从15.2%提升至22.9%。在2026年一季度,智谱API调用定价提升83%,调用量反而增长400%,供不应求。

银河通用与智谱站在AGI的两条岔路口上,一个用代码构建数字世界的智能,一个用钢铁和电机再造物理世界的智能。两条路径看似平行,却在一个更深层的结构上紧密耦合。

02 技术跃迁 从“控制机器”到“理解世界”

2026年央视春晚的贺岁微电影中,一个轮式双臂机器人完成了盘核桃、捡玻璃碎片、叠衣服、串烤肠等一系列高难度操作。这个机器人名为Galbot,驱动它的是银河通用自研的“银河星脑”(AstraBrain),全球首个集成“大脑-小脑-神经控制”于一模的端到端具身大模型。

要知道,过去传统机器人系统采用“感知-规划-控制”三级分离架构,每一层由不同团队用不同技术路线完成,延迟和误差逐层累积。银河星脑则将这三层打通为一条神经网络,从多模态感知到实时反馈控制全链路端到端训练,实现了“边看、边想、边做”的实时决策。

在“世界-动作模型”(World-Action Model,WAM)这一核心技术路线上,银河通用走在了全球最前沿。

早在2025年,银河通用团队在计算机视觉顶会ICCV上首次提出WAM概念,它将Google的VLA(视觉-语言-动作模型)与OpenAI Sora代表的世界模型两条技术路线统一起来。而这一技术路线的战略判断,与全球顶尖科技企业对具身智能终局的思考不谋而合。英伟达研究科学家Jim Fan在2026年接受采访时直言:WAM是“robotics endgame”(机器人技术的终局)。

在“专”的层面,银河通用打出了两记全球首创的重拳。

一是全球首个全自主人形机器人与真人网球对抗系统。2026年3月,一段人形机器人在网球场上与真人选手连续对拉的视频在全球科技圈刷屏。机器人在毫秒级内完成来球感知、轨迹预判、跑位调整和挥拍击球的全闭环。

特斯拉对于这段视频,海外业内顶尖人士集体给出极高评价:特斯拉CEO马斯克在Twitter上直呼“Insane”;AI知名评论员Andrej Karpathy发表评论表示惊叹,一度“怀疑”视频是由AI生成;AI评论员Andrew Kang更是直言,AlphaGo时刻已经到来!这些评论足以印证本次具身智能突破的划时代价值。

AI研究者Andrej Karpathy第一次看到视频时,认为这是AI视频生成软件制作的假视频,他无法相信机器人的运动已经真实到这种程度。

背后的算法核心是银河通用提出的LATENT算法。它通过构建“运动技能空间”与隐空间动作屏障,让机器人从不完美的人类动作数据中自主学习高动态技能,而不是依赖远程遥控或预编程。

二是灵巧手“转笔”,全球唯一实现从仿真到真机迁移的高精度灵巧操作。英伟达早在2023年就在仿真器中展示了灵巧手转笔,但一直无法在真实世界中复现。银河通用提出的灵巧手神经动力学模型(DexNDM),通过在仿真器中构建逼真的碰撞模型,再用真实数据训练“灵巧手世界模型”弥合虚实差距,最终在全球范围内首次实现了真实机器人的手指转笔。

2026年4月,银河通用推出了AstraBrain WAM 0.5,全球首次实现虚实共融、人机混合、质量参差、有无动作标签数据的统一有效利用。同期发布的通用小脑AstraBrain-WBC 0.5,基于2万小时人类动作数据训练,在全身闭环控制与实时指令跟随方面达到行业领先水平。

从“AlphaGo时刻”到“ChatGPT时刻”,数字AI用了十年走完从专到通的路。而具身智能正以加速度进行同样的演进。

王鹤的判断是:当预训练模型在人类无需专门后训练就能完成的技能上达到70%到80%成功率时,具身智能的“ChatGPT时刻”就会到来。

如果技术突破是银河通用的“上半场”,那么商业化落地就是它正在书写的“下半场”。

2026年6月25日,新京报披露了一个里程碑事件:银河通用Galbot S1具身智能重载机器人已在宁德时代量产线上7×24小时工作超过3个月,这是全球首次实现人形机器人在新能源智能制造生产场景的“常态化自主作业”。

Galbot S1以双臂50公斤级重载能力、纯视觉柔性定位与360度全向避障为核心优势,在宁德时代产线上承担模组与电池包生产环节的物料转运等长程自主任务。它采用纯视觉方案,无需依赖二维码或反光板定位,即可在复杂工厂环境中完成操作,这意味着它真正具备了对非结构化环境的适应能力。

与宁德时代的合作只是工业落地的起点。截至目前,银河通用的工业人形机器人已进入宁德时代、博世集团、西门子、丰田汽车、现代汽车、北汽集团、上汽集团、极氪汽车、长城汽车等制造企业的真实生产线,累计订单达数千台。

在零售领域,银河通用同样跑通了规模化的路径。搭载银河星脑的机器人已在近40家智慧零售药店和170家“银河太空舱”零售店中实现7×24小时自主卖货运营,覆盖全国多个城市核心商圈。

医疗领域是银河通用的第三个主战场,公司与四川大学华西第二医院合作共建具身智能医疗机器人研发平台,探索机器人在医疗康养场景中的应用。

回看银河通用的商业化策略,一条清晰的路径是:它并不追风口、抢量产规模,而是面对终端客户,银河通用执行严选科技策略,对SR(系统需求)要求极其严,每个场景都经过技术可行性、工程可交付性与商业可持续性的三重验证。

公司不追求"万物皆可做",而是聚焦三大场景,在工业场景用高难度命题锤炼技术壁垒;零售场景在动态环境中检验机器人的通用性与泛化能力;医疗康养场景将已验证的技术框架向长尾市场渐进延展。这套路径的背后有一个核心判断,量产数量从来不是定义技术成功的唯一标尺。

具身智能的历史性拐点已经到来,机器人正在从“执行工具”进化为“自主系统”。

03 结尾

如果将AGI看作一个完整的智能系统,智谱提供的是数字世界的AI,让模型学会理解语言、规划任务、推理逻辑;银河通用提供的是物理世界的AI,让机器人学会感知物理世界、精细操作、自主运动。

智谱的商业模式建立在“Token经济”之上:每一次API调用、每一次模型推理,都在将智能转化为可计量的经济价值。而银河通用的商业逻辑建立在“物理经济”之上:每一台部署在产线上的机器人,都在直接创造物理世界的生产效率。

两种经济模式并非孤立,而是互为前提。未来的通用机器人需要“大脑”来理解和规划任务,也需要“身体”来执行和反馈。

北京AI产业的“双龙头”格局,不是在同一个赛道里选冠军,而是在两条赛道里各有一个“决策龙头”。从清华KEG实验室的代码到港股万亿市值,北大EPIC Lab的仿真到宁德时代产线上的真机,从实验室到万亿市场,从论文到产线,中国AI正在形成一个全球罕见的“数字+物理”双轨并行的产业闭环。

北京市市长殷勇在调研时强调,要“统筹推进基础大模型与具身智能协同攻关,聚焦关键领域集中突破,筑牢软硬件技术底座”。

这句话点出了北京AI产业布局的结构性优势。全国最丰富的人工智能创新资源、最具活力的产业生态、最坚实的新型基础设施,这些要素的叠加,让北京成为全球少数同时布局“数字智能”与“物理智能”两大AGI路径的城市之一。

海淀区2000多家人工智能企业和330多家具身智能企业,构成了一个多层次、多梯队的创新矩阵。而智谱与银河通用,就是这个矩阵中最耀眼的双子星。

注:文/白华,文章来源:创业最前线(公众号ID:chuangyezuiqianxian),本文为作者独立观点,不代表亿邦动力立场。

文章来源:创业最前线

广告
微信
朋友圈

FAQ回顾

通用人工智能有哪些主流技术发展路径?

通用人工智能主要有两大技术发展路径:一是以智谱为代表的“数字大脑”路径,聚焦大模型研发,实现长程任务处理、逻辑推理等数字智能能力;二是以银河通用为代表的“物理身体”路径,深耕具身智能领域,研发可在真实场景自主作业的通用机器人。

具身智能的商业化落地场景主要有哪些?

目前具身智能商业化落地主要聚焦三大核心场景:一是工业制造场景,可在新能源、汽车等产线承担物料转运等长程自主任务;二是智慧零售场景,可实现7×24小时自主卖货运营;三是医疗康养场景,可探索适配医疗护理需求的智能化服务。

智谱GLM-5.2大模型有什么核心技术优势?

GLM-5.2是智谱2026年上线开源的新一代旗舰大模型,主攻“长程任务”,可支撑AI连续工作数小时自主完成完整大型工程,在全球百万用户参与盲测的Code Arena榜单上位列全球可用模型第一。

具身智能的ChatGPT时刻什么时候会到来?

银河通用创始人王鹤判断,当预训练模型在人类无需专门后训练就能完成的技能上达到70%到80%成功率时,具身智能的“ChatGPT时刻”就会到来。

这么好看,分享一下?

朋友圈 分享

APP内打开

+1
+1
微信好友 朋友圈 新浪微博 QQ空间
关闭
收藏成功
发送
/140 0