软银集团下周发布的第三季度财报将是这家日本科技巨头回应有关其正考虑增加对OpenAI投资报道的 首个机会。据行业研究称，将其目前持有的11%股份进一步扩大可能会对其信用评级产生压力，因为据 报道300亿美元的投资增加可能导致该集团的贷值比触及35%的降级触发点，这可能需要至少150亿美元 的资产出售和保证金贷款才能将该比例维持在25%以下。继今日晚些时候丰田汽车之后，日产汽车和本 田技研工业是下一批公布财报的车企，由于三家公司均拥有美国市场敞口，日元走软被认为对它们有 利。 本文源自：金融界AI电报 ...

Core Insights - OpenAI and Anthropic have both launched new AI models focusing on programming, autonomous collaboration, and enterprise applications, indicating a competitive landscape in the AI sector [1] Group 1: OpenAI Developments - OpenAI introduced the GPT-5.3-Codex model, which enhances programming performance and reasoning capabilities, with a core reasoning speed improvement of 25% compared to the previous generation [1] - The company also launched the OpenAI Frontier platform, designed to integrate disparate enterprise systems and data, facilitating one-stop management, deployment, and construction of AI agents for business processes [1] Group 2: Anthropic Developments - Anthropic released the Claude Opus 4.6 model, which supports a context window of 1 million tokens, a fivefold increase from the previous 200,000 tokens [1] - The model features breakthroughs in programming, task planning, and code safety, with a new "Agent team" function that automates the breakdown and allocation of complex tasks among multiple agents, enhancing task processing efficiency [1]

谷歌又交出一份令人咂舌的财报。但最近市场显得有点焦虑，并没有给予足够热切的回应。 截至2025年12月，谷歌母公司Alphabet在第四季度实现1138亿美元营收，全年更是首次突破4000亿美元。其中，谷歌云第四季度营收达到176.6亿美元，同 比增长48%，显著高于分析师预期。谷歌云的年化运营收入，已经超过700亿美元。这样的好日子还将继续，谷歌的订单积压环比增长了55%，至2400亿 美元。 此外： 2000万亿token消耗门槛 如果假定面向企业与消费者的token消耗以相近的速度增长，那么，截至去年12月的上一财季，谷歌每月消耗token数量达到了1820万亿tokens。作为对 照，上个月，旗舰豆包大模型升级至Seed-1.8，直指"通用现实世界智能体"，将字节跳动日均token消耗提升到63万亿量级，也即每月消耗1890万亿 tokens。 据此，在推理token消耗规模这一指标上，近一年来，中美头部公司已经进入同一数量级竞争区间。这也是为什么在达沃斯世界经济论坛上，DeepMind创 始人哈萨比斯（Demis Hassabis）中国公司极其擅长追赶，尤其是特别表扬了字节跳动。 百亿美元收入门槛 ...

Group 1: Market Overview - The recent decline in various asset classes, including U.S. stocks, gold, and Bitcoin, is attributed to a clear logic of deleveraging and market corrections [3][4][6] - Bitcoin's sharp drop is primarily due to investors using leverage, leading to forced liquidations as prices fell, creating a domino effect [3] - Gold and silver prices have also decreased, influenced by reduced geopolitical tensions and a stronger U.S. dollar, alongside selling pressure from investors needing to cover margin calls [3][4] Group 2: U.S. Stock Market Dynamics - The Dow Jones Industrial Average fell nearly 600 points, with the S&P 500 erasing its gains for the year, driven by concerns over weak employment data and a potential shift in Federal Reserve interest rate expectations [4] - The market is undergoing a repricing phase, reflecting uncertainty about economic conditions and interest rate policies [4] Group 3: Technology Sector Insights - The technology sector is experiencing a divergence, with companies like OpenAI advancing AI applications while others, such as those involved in traditional finance, face challenges [4][6] - The release of AI models capable of financial research has raised concerns about the impact on traditional financial services, indicating a potential industry transformation [6] Group 4: Investment Strategies - Investors are advised to avoid leverage, particularly in volatile assets like cryptocurrencies, and to focus on long-term, stable investment strategies [5] - It is recommended to refrain from chasing market trends or attempting to time the market during periods of panic, instead waiting for clearer signals before making investment decisions [5] - Diversification across asset classes, including U.S. stocks and precious metals, is emphasized to mitigate risks associated with market fluctuations [5] Group 5: Macro Trends and Policy Implications - Global uncertainties are increasing, influenced by geopolitical developments and policy changes, which can subtly affect market dynamics [5] - Investors should align their strategies with broader policy trends and industry developments rather than acting against prevailing market sentiments [5]

Core Insights - The article discusses the rapid evolution of Agentic Memory in AI, emphasizing that without memory, AI agents are merely advanced autocomplete tools. To handle complex projects or long-term tasks, AI must possess a structured long-term memory mechanism [1][3]. Industry Trends - There is a significant shift in the AI industry towards persistent memory as a critical capability for agents, with major players like OpenAI and Anthropic pushing the limits of context windows [1][3]. - Sequoia Capital highlights that one of the core challenges for future agents is achieving persistent identity, which involves remembering user interactions over time while maintaining consistent understanding and context [3]. Company Highlights - Feeling AI, a newly established startup, has recently launched MemBrain1.0, achieving state-of-the-art (SOTA) results in several memory benchmarks, surpassing existing systems like MemOS and EverMemOS [4][5]. - The team behind Feeling AI is led by Dr. Dai Bo, a young scientist in generative AI, who has previously worked at NTU and Shanghai AI Lab. The team has completed two rounds of funding exceeding 100 million yuan, focusing on world models and 3D dynamic interaction [4][19]. MemBrain1.0 Performance - MemBrain1.0 achieved new SOTA with accuracy rates of 93.25% and 84.6% on the LoCoMo and LongMemEval benchmarks, respectively, due to its refined entity-time context management design [9]. - In the PersonaMem-v2 benchmark, MemBrain1.0 outperformed existing methods with an accuracy of 51.50%, demonstrating deep insights into user preferences [10]. Technical Innovations - MemBrain's architecture allows for flexible deployment and asynchronous memory updates, enhancing the system's adaptability [15]. - The system's design focuses on precise extraction of entities and timestamps, which is crucial for high-level tasks like associative analysis and logical reasoning [16]. - MemBrain organizes information into semantic units that can be loaded on demand, allowing for deeper participation of large language models (LLMs) in reasoning tasks [17]. Future Outlook - The article suggests that the ability to solve the "forgetting syndrome" of agents will be key to advancing towards Artificial General Intelligence (AGI) [27]. - The evolution of memory capabilities is seen as a transition from "stateless" single calls to "conscious" continuous improvement, marking a new beginning for AI in user interaction and co-creation [27].

这就非常有意思了。 就在今天（2026年2月6日），Anthropic和OpenAI跟约好了一样，短短20分钟里先后扔出了两枚深水炸 弹，那就是，Claude Opus 4.6和GPT-5.3 Codex。 要把两家的发布拆成「它们到底在强化啥能力」来看，差别其实特别明显。 没有预热、没有铺垫，甚至没有留给市场消化的时间。这是一场极其明确的 「模型遭遇战」。 01 这场正面硬刚，清清楚楚标志着AI竞赛，已经从过去两年的「聊天对话能力竞赛」，正式切换到了一 个全新阶段，Agent自治时代。 说重点，Anthropic那边，核心三个点：可控的Agent行为边界、任务拆解和执行的稳定性，还有更明确 的人类监督和回滚机制。 它们没把Agent吹成「什么都能干」的全能选手，重点就强调一点：Agent必须在清晰的规则、权限和审 计框架下运行。 而OpenAI给出的方向，则明显更激进： 更强的自主规划能力、多步骤跨工具的持续执行，还有模型要对复杂目标负全责。说白了，信号很直 接，Agent能接管一整个任务流程，从看懂目标到最后交出结果。 总结一句就是，Anthropic在强调「别出错」，OpenAI在赌「先跑起来」。 这就 ...

1. SpaceX 100万颗卫星申请获受理，马斯克否认将推出星链手机 2. 李斌兑现承诺！蔚来：2025年Q4经营利润至少7个亿 3.董宇辉卖春联被骂上热搜！网友怒斥：这种钱你也赚 要闻提示 NEWS REMIND 4.阿里巴巴大模型品牌统一为千问，千问APP启动30亿免单 今日头条 HEADLINE NEWS SpaceX 100万颗卫星申请获受理，马斯克否认将推出星链手机 2月6日消息，日前，美国联邦通信委员会表示，已受理SpaceX的数据中心申请。此前，外媒称，SpaceX 申请发射多达100万颗卫星，以建立一个环绕地球的轨道数据中心网络。1月30日，SpaceX向美国联邦通 信委员会提交了申请，将该项目描述为"一个拥有前所未有的计算能力的卫星星座，可为先进的人工智能 （AI）模型及其相关应用提供支持"。报道称，该计划规模惊人，远超马斯克现有的"星链"星座，后者目 前在地球轨道上拥有超过9600颗卫星。马斯克在社交平台X上转发了关于SpaceX申请发射多达100万颗 卫星的报道。 值得注意的是，日前，有消息传出SpaceX正规划推出星链专属智能手机，可直接连接其星链(Starlink)卫 星星座。马斯 ...

5. 蔚来：预计2025年第四季度录得经调整经营利润7-12亿元，为首次实现单季度盈利，相比2024年第四 季度55.44亿元的经调整经营亏损有显著改善。 // 大中华地区公司要闻 // 1. 百度集团：董事会已批准50亿美元股份回购计划及首次股息政策，预计2026年首次发放股息，旨在提 升股东回报，显示公司对自身发展前景的信心及对投资者的重视。 2. 台积电：计划投资约170亿美元在日本熊本厂量产3纳米芯片，同时有消息称公司正评估调整熊本第二 座晶圆厂的制程规划，引发市场对其资本支出节奏的关注。 // 热点头条 // 1. 美团：于2026年2月5日订立股份转让协议，同意以7.17亿美元收购叮咚买菜母公司DINGDONG FRESH HOLDING LIMITED全部已发行股份，收购完成后目标公司将成为美团间接全资附属公司。叮 咚买菜CEO梁昌霖表示，合并后三大核心竞争力将在更大平台发挥价值，业务和团队保持稳定。 2. 英伟达：因存储芯片短缺，今年将不会推出新款游戏图形芯片，为三十年来首次，同时削减当前一代 GeForce RTX50系列显卡产量。作为全球芯片巨头，此举可能影响整个游戏硬件产业链及消费者升级 ...

AI音频赛道，冲出一家百亿美元独角兽！ 智东西2月5日报道，昨天，英国AI音频独角兽ElevenLabs宣布完成5亿美元（约合人民币34.7亿元）的D轮融资，估值达110亿美元（约合人民 币763.5亿元）。其估值较去年年初的33亿美元，实现了超230%的飞速增长。ElevenLabs联合创始人兼CEO Mati Staniszewski还透露，该公司 已在考虑IPO事宜。 Mati Staniszewski发文宣布融资（图源：X平台） ElevenLabs于2022年成立于英国伦敦，最初主要从事文本转语音模型的开发，后来也陆续打造了语音转文本模型、AI音效模型、AI配音模型、 AI音乐模型等领域。 该公司既向企业提供语音API服务，也为创作者和品牌提供音频生成平台ElevenCreative，还通过ElevenAgents平台向企业提供AI语音客服服 务。你甚至能在知名游戏《堡垒之夜》里找到由ElevenLabs生成的音频。2025年底，ElevenLabs的ARR（年度经常性收入）已经超过3.3亿美 元（约合人民币22.9亿元）。 本轮融资由红杉资本领投。已参与ElevenLabs多轮融资的a16z追加4 ...

炒股就看金麒麟分析师研报，权威，专业，及时，全面，助您挖掘潜力主题机会！ 来源：人工智能顾问 日前，国内权威大模型评测机构SuperCLUE正式发布《2025年年度中文大模型基准测评报告》。这份被 誉为"AI界年终成绩单"的重磅榜单，再次点燃了行业关注。 这标志着国产大模型正从"追赶者"迈向"并行者"，甚至在特定领域实现"领跑"。 01 海外闭源模型领跑，国产模型紧咬不放 本次测评从数学推理、科学理解、代码生成等六大核心维度，全面检验大模型的综合智能水平。最终综 合得分排名中： ●第1名：Claude-Opus-4.5-Reasoning（Anthropic）—68.25分 ●第2名：Gemini-3-Pro-Preview（谷歌）—65.59分 ●第3名：GPT-5.2（high）（OpenAI）—64.32分 三大海外闭源模型凭借其强大的工程优化与数据积累，继续领跑全球。 但值得注意的是，第4名由国产开源模型Kimi-K2.5-Thinking以61.50分强势拿下，成为综合排名最高的 国产模型；而阿里云推出的高性能模型Qwen3-Max-Thinking则以60.61分位列第6，紧随其后，展现出中 国头 ...