自驾行业今年还是很精彩的，在整体下沉的关键节点，都很卷。卷技术、卷成本、卷效率。我们今年亦是如此，扩充了很多 B端的客户，也开始尝试从线上走向线下。C端也慢慢从普适性的能容逐渐专业化和精细化。 上半年不少自驾的同学转行去了具身，包括现在也是如此，L4/具身/无人机几个行业在大批量招人，而自驾又是相对成熟的 AI领域，所以自驾的算法人才非常受欢迎，几个头部企业的薪资很到位（大疆/宇树/智元/哈啰等等）。 下周就要迎来26年了，也到了年末盘点的时候。 搞过自驾的人，用过大集群，解过各种corner case，上下游协同能力强，这些都是其他几个行业所欠缺的。 今年，自驾的头部技术收敛到几个大方向上：一段式端到端、VLA、世界模型（重建+仿真）、强化学习。我们接触到的中 游厂商还在攻坚OCC、无图、多传感器融合感知等等，明年这些公司都有大量hc开放。 今年，自动驾驶之心的付费社区的成员正式突破4000人了。如果想看技术路线的发展、各类圆桌、研报、职位信息，可以多 来逛逛。 新的一年，也感谢新老粉丝的支持，我们为大家推出了众多福利优惠。新的一年大家再接再厉。 星球新人六折券，续费五折券 欢迎添加助理咨询活动 ...

Core Insights - The AI industry is transitioning from a focus on model performance to a comprehensive competition involving technology systems, business paths, infrastructure, and ecosystem building for 2026 [4][12]. Group 1: Major Players and Competitive Landscape - Google has established a significant user mindshare barrier in multimodal tasks with its Gemini model, despite ChatGPT being preferred for text-based interactions [6][7]. - OpenAI may experience a rebound in 2026 as supply chain issues are resolved, potentially leading to increased user engagement and product capabilities [13][14]. - Anthropic is positioned as a strong player in the enterprise AI market, focusing on B2B applications and addressing pain points more effectively than competitors [15][16]. - Meta is projected to achieve an annual AI revenue scale of $60 billion, benefiting from improved advertising efficiency due to AI applications [18][20]. Group 2: Technological Developments and Trends - The World Model is seen as a critical differentiator in the next generation of AI technology, with companies like Meta exploring human-like evolution in AI understanding [28][31]. - The competition for AI application entry points is intensifying between operating system providers and app developers, with both sides facing unique challenges [32][34]. - The development of edge AI is driven by user demands for data sovereignty and privacy, leading to increased hardware requirements for local processing [40][41]. Group 3: Infrastructure and Bottlenecks - Optical communication and interconnect technologies are expected to see explosive growth, with Google’s Optical Circuit Switching technology being a key focus [48]. - Storage is transitioning from a cyclical to a growth trend, driven by enterprise AI demands and the need for extensive data retention [49][52]. - Power consumption is becoming a significant bottleneck for AI development, with the need for efficient energy solutions becoming critical as demand increases [53][54]. Group 4: Market Applications and Future Outlook - Enterprise AI is anticipated to penetrate various sectors, including finance and HR, with tangible products expected to emerge by 2026 [55][60]. - The integration of AI into prediction markets may shift the focus from gambling to rational risk hedging, enhancing decision-making capabilities [61][63]. - The Agent model is expected to proliferate in payment automation and e-commerce, streamlining operations across platforms [64].

最近我们 复盘 了去年「2025 AI Best Ideas」提出的 20 个关键预测，发现绝大部分关于技术方向与 格局演化的 AI 预测已经兑现。而站在当下看 2026 年这个关键时间节点，市场已经显现出了更明显 的分歧：Gemini 3 发布后，Google 能否保持长期领先？OpenAI 是否有机会在 2026 年实现逆转？在 AI 入口竞争中，是操作系统占优，还是超级 APP 更具潜力？ 因此我们组织了一场「2026 AI Best Ideas」社群讨论，AI researchers、创业者、产品经理和一二级 投资人围绕 2026 年 AI 公司竞争格局、AI 应用与 Agent 形态、算力与 infra 瓶颈，以及 AI 在具体 行业中的落地路径等关键问题，展开了一次深入的讨论。 本篇文章并不是一份单一视角的年度判断，而是来自拾象 Best Ideas 社群集体讨论的精华开源。我 们希望它不仅是一份年度预测，更能帮助读者理解：AI 是一次真实且长期的生产力革命，在模型 厂商交替领先的格局中，真正的赢家不仅要关注技术实力，更要在高度不确定的环境中实现长期价 值。 ⬇️ 滑动或点击查看大图 ⬇️ 讨论主 ...

点击下方 卡片 ，关注" 自动驾驶之心 "公众号 戳我-> 领取 自动驾驶近30个 方向 学习 路线 >>自动驾驶前沿信息获取 → 自动驾驶之心知识星球 最近自动驾驶的两大前沿方向：VLA和世界模型，已经有明显的融合趋势 。这一想法是十月份看到中科院的 DriveVLA-W0，因此笔者借这个机会分别调研了 VLA 和 World Model 相关的工作，并且思考一下 这二者结合 的可能性。 太长不看版： VLA和世界模型并不冲突，终极目标是一致的。世界模型可以作为数据引擎、闭环引擎，甚至可以参与到VLA 的模型训练过程中，融合是大趋势，落地是我全都要。 经过几周的调研、分析，有了些成果和自己的心得，所以也想理一理，分享给自动驾驶之心的小伙伴们，主 要分为以下几个部分： 输入端：融合多模态感知 VLA的输入整合了视觉、传感器与语言等多模态的信息。核心视觉输入通过多摄像 头图像生成BEV或体素表征，以理解空间结构；传感器（如激光雷达、毫米波雷达）提供几何与动态补充； 语言输入则是关键创新，支持导航指令、交互问答与规则描述，使系统能理解人类意图与常识，构建出超越 传统纯视觉感知的环境理解。 自动驾驶技术诞生到发展至 ...

今年前三个季度，国内20万元以上乘用车市场份额占比30%，13万元以下市场份额则高达50%，但后者 多数车型尚未配备城区辅助驾驶功能。这一广阔的蓝海市场，正吸引着地平线、Momenta等智驾厂商加 速布局，全力抢占市场先机。 今年4月，地平线正式推出基于征程6系列芯片的城区辅助驾驶解决方案——HSD（Horizon SuperDrive）。尽管并非该赛道的先行者，但地平线已快速迈入大规模量产阶段。11月，随着星途ET5 正式上市，地平线的HSD解决方案同步实现量产；另一款搭载该方案的车型深蓝L06也于同期发售。两 款车型上市短短两周后，地平线HSD的激活量便突破12000辆，量产落地成效显著。 除了推出全新的解决方案，地平线还通过生态拓展加速市场渗透。12月初的地平线技术生态大会上，公 司公布了两大生态推进举措：一是拓展生态合作模式，新增算法服务模式"HSD Together"，并已与日本 电装、大众的合资公司CARIZON（酷睿程）、HCT（智驾大陆）达成合作；二是引入更多生态合作伙 伴，元戎启行、卓驭等企业已加入其生态体系。 缺乏芯片研发能力的算法公司、软硬研发实力薄弱的车企，正纷纷向地平线聚拢。地平线接 ...

Core Insights - GAIA-3 represents a significant advancement in the evaluation of autonomous driving systems, transitioning world modeling from a visual synthesis tool to a foundational element for safety assessment [4][20] - The model combines the realism of real-world data with the controllability of simulations, enabling the generation of structured and purposeful driving scenarios for safety validation [6][20] Group 1: GAIA-3 Features - GAIA-3 is a powerful testing tool that can modify vehicle trajectories, weather conditions, and adapt to different sensor configurations [3] - It is built on a latent diffusion model with 15 billion parameters, doubling the video tokenizer size compared to its predecessor GAIA-2 [3][19] - The model allows for the generation of controlled variants of real-world driving sequences, maintaining consistency in the environment while altering vehicle behavior [6][8] Group 2: Safety and Evaluation - GAIA-3 addresses the limitations of traditional testing methods by generating systematic variations of critical safety scenarios, such as collisions, using real-world data metrics [7][8] - The model enables offline evaluation of autonomous systems by recreating unexpected events, allowing for quantitative testing of recovery capabilities in edge cases [9][20] - It emphasizes consistency in generated scenarios, ensuring that changes in vehicle behavior do not disrupt the physical and visual coherence of the environment [8][11] Group 3: Data Enrichment and Robustness - GAIA-3 enhances data coverage by generating structured variants from rare failure modes, facilitating targeted testing and retraining [12][13] - The model supports controlled visual diversity, allowing for measurable changes in appearance while keeping the underlying structure consistent, thus improving robustness assessments [11] - It can transfer scenarios across different sensor configurations, enabling data reuse across various vehicle projects without the need for paired collection [10] Group 4: Technical Advancements - The advancements in GAIA-3 are driven by increased scale, with training compute five times that of GAIA-2 and a dataset covering eight countries across three continents [16][19] - The model captures critical spatial and temporal structures, enhancing the fidelity of generated scenarios and improving the understanding of causal relationships in driving behavior [19][18] - GAIA-3's capabilities provide a reliable framework for structured, repeatable testing, marking a significant step towards scalable evaluation of end-to-end driving systems [20]

December 14, 2025 09:00 PM GMT The Robot Almanac Vol. 2: How to Train Your Robot; Geopolitics; Rare Earths; Sagan's Prophecy Morgan Stanley Global Embodied AI Team December 2025 The content addressing private companies is being provided for informational purposes only and does not constitute a solicitation or imply future research coverage if the company goes public. Content is based on unaudited information. No investment recommendation is provided as there is limited public information available for priva ...

鹭羽 发自 凹非寺 量子位 | 公众号 QbitAI 世界模型赛道，又有老面孔新鲜入局！ 就在刚刚，Runway发布旗下首个通用世界模型 GWM-1 。 不止于此，还打包发布了一系列世界模型变体： 而这些通通都是基于最新版 Gen-4.5 建立的。 是的！Runway这次还把Gen-4.5来了个大升级。 模拟真实环境的GWM Worlds； 模拟人物对话的GWM Avatars； 模拟机器人操作的GWM Robotics。 …… 看来年末大促销的不只有圣诞老人奥特曼，还有好莱坞名导Runway。 话不多说，上实机： 世界模型全家桶发布 根据官方介绍，GWM-1是基于Gen-4.5构建的，这是Runway最新的视频生成模型。 但和Gen-4.5有所不同的是，GWM-1采用的是 自回归 架构，它可以根据之前的记忆内容，进行逐帧预测生成。 另外模型支持实时交互控制，包括调整相机姿态、修改机器人操作指令或音频。 它目前包含三个变体： 1、GWM Worlds：用于实时环境的模拟与探索。 GWM Worlds能够让用户在连贯、有反应的世界中自由移动，而无需手动设计每个空间。 具体来说，用户首先需要为模型提供一个可供参考 ...

Financial Data and Key Metrics Changes - In Q3 2025, the company reported revenue of $25.4 million, a growth of 72% year-over-year [44] - Gross profit margin improved significantly from 9.2% in Q3 2024 to 18.4% in Q3 2025, with gross profit of $4.7 million [50] - Net loss for Q3 was $61.6 million, compared to $42.1 million in the same period last year [54] Business Line Data and Key Metrics Changes - Robotaxi services revenue reached $6.7 million, representing a growth of 89.5% year-over-year and 338.7% quarter-over-quarter [45] - Fare charging revenue surged by 233.3%, driven by increased user adoption and operational efficiency [46] - Robot truck service revenues were $10.2 million, growing by 8.7% [49] Market Data and Key Metrics Changes - The company expanded its robotaxi footprint to eight countries globally, indicating strong international growth potential [47] - The daily net revenue per vehicle reached CNY 299, with an average of 23 orders per day [51][76] - The total number of registered users nearly doubled within a week of launching the Gen-7 Robotaxi [10] Company Strategy and Development Direction - The company aims to scale its fleet to over 3,000 vehicles by 2026, leveraging the momentum from the recent Hong Kong IPO [57] - The launch of the Gen-7 Robotaxi has validated the business model, allowing for deeper collaborations and operational expansion in Tier 1 cities [64] - The company is focusing on technological innovation and operational efficiency to enhance its competitive edge in the autonomous mobility sector [22] Management's Comments on Operating Environment and Future Outlook - Management expressed confidence in sustaining robust growth momentum, driven by fleet expansion and improved user experience [62] - The successful Hong Kong IPO is expected to accelerate R&D investments and solidify the company's technology leadership [57] - The company views the entry of new players into the robotaxi market as a positive sign of growing recognition and potential for large-scale commercialization [85] Other Important Information - The company completed a dual primary listing on the Hong Kong Stock Exchange, raising over $800 million [4] - The Gen-7 Robotaxi has achieved city-wide unit economic break-even in Guangzhou, validating the business model [8] - The company is transitioning to a satellite model for fleet expansion, allowing for greater capital efficiency [58] Q&A Session Summary Question: Updates on fleet size and outlook for 2026 - Management expects to outperform the target of 1,000 robotaxis by year-end and aims for over 3,000 vehicles in 2026, driven by user experience and fleet density [62] Question: Outlook for fare charging revenues - Fare charging revenue surged by 233%, with expectations for sustained growth as fleet expansion continues [67][71] Question: Assumptions behind the unit economic break-even - The daily net revenue per vehicle is CNY 299, with 23 average orders per day, supported by operational cost management [76][78] Question: Views on new entrants in the robotaxi space - The company sees new entrants as a positive sign but highlights significant barriers to entry, including business, regulatory, and technical challenges [85][88] Question: Factors behind faster expansion of operational areas - The company attributes faster expansion to the number of robotaxi vehicles and the inherent generalization capabilities of its technology stack [100][101]

Financial Data and Key Metrics Changes - In Q3 2025, the company reported revenue of $25.4 million, a growth of 72% year-over-year [44] - Robotaxi services revenue reached $6.7 million, representing a growth of 89.5% year-over-year and 338.7% quarter-over-quarter [45] - Gross profit margin improved from 9.2% in Q3 2024 to 18.4% in Q3 2025, with gross profit of $4.7 million [48] - Net loss for Q3 was $61.6 million, compared to $42.1 million in the same period last year [50] Business Line Data and Key Metrics Changes - Robotaxi revenue surged by 90% year-over-year, with fare charging revenues growing over 200% year-over-year [12] - Robot truck service revenues were $10.2 million, growing by 8.7% [47] - Licensing and application revenues were $8.6 million, growing significantly by 354.6% [47] Market Data and Key Metrics Changes - The company has established a robotaxi presence in eight countries, including new markets like Qatar [17] - Daily net revenue per vehicle reached CNY 299, with an average of 23 orders per day [49] - The total number of registered users nearly doubled within a week of launching Gen7 [10] Company Strategy and Development Direction - The company aims to expand its fleet to over 3,000 vehicles by 2026, leveraging the satellite model for fleet expansion [56] - The recent Hong Kong IPO raised over $800 million, strengthening the balance sheet for mass production and commercialization [4][52] - The focus is on technological innovation and creating lasting value through efficient autonomous mobility services [22] Management's Comments on Operating Environment and Future Outlook - Management expressed confidence in scaling operations following the city-wide unit economic break-even milestone achieved in Guangzhou [66] - The company sees increasing recognition and confidence in the robotaxi industry's potential for large-scale commercialization [72] - Future growth will be supported by partnerships with local governments and third-party operators [58][97] Other Important Information - The company has ramped up production, with over 600 Gen7 Robotaxis produced by November, exceeding the full-year target of 1,000 vehicles [11] - The Gen7 Robotaxi has achieved city-level unit economics break-even shortly after launch, validating the business model [8] Q&A Session Summary Question: Updates on fleet size and outlook for 2026 - Management expects to outperform the target of 1,000 robotaxis by year-end and aims for over 3,000 vehicles in 2026, driven by the Gen7 launch [56] Question: Outlook for fare charging revenues - Fare charging revenue surged 233% in Q3, driven by user demand and operational optimizations, with expectations for sustained growth as fleet expands [61] Question: Assumptions behind the unit economic break-even - Daily net revenue per vehicle is CNY 299, with 23 orders per day, supported by operational cost management and hardware depreciation strategies [67] Question: Views on new entrants in the robotaxi space - The company sees new entrants as a positive sign of growing confidence in the industry, but emphasizes the challenges of business, regulatory, and technical hurdles [72][74] Question: Factors behind faster expansion of operational areas - The company attributes faster expansion to the number of robotaxi vehicles and the ability to handle corner cases effectively [82]