市场层面，根据《中国大模型中标项目监测与洞察报告（2025年Q2》，科大讯飞中标项目数量、中标 金额在2025年第二季度实现双第一。（文猛） 新浪科技讯 7月10日晚间消息，科大讯飞发布2025年半年度业绩预告，预计实现营业收入同比增长 15%-20%，归母净利润同比增长30%-50%，扣非净利润增长17%-34%。报告期内，公司经营活动现金流 净额同比提升超7亿元，增幅约50%；销售回款总额达103亿元，同比增长13亿元。 责任编辑：何俊熹 科大讯飞称，公司踏实推进长期、可持续、高质量发展，并保持了经营的良性健康发展。业绩增长主要 源于公司积极抢抓通用人工智能的历史新机遇，继续保持面向未来抢占人工智能根红利的战略投入，研 发投入占营业收入比重约20%；同时，源头技术驱动的战略布局成果不断显现，核心业务呈现出健康发 展的势头，特别是TO C业务继续保持快速增长。 报告期内，基于"讯飞星火大模型"的AI硬件在"618"电商节销量同比增长42%；其中，科大讯飞AI学习 机连续三年位居京东及天猫"618"全周期销售额榜首，2025年半年度收入保持翻番增长。 此外，公司4月20日再次升级的讯飞深度推理70B大模型星火 ...

作为引领新一轮科技革命和产业变革的战略性技术，人工智能（AI）应用场景广阔，市场空间巨大， 深刻改变着人类生产生活方式。从手机应用、智能音箱等智能助手，到机械臂、人形机器人等具身智 能，再到能够提供个性化服务的智能体以及更多AI原生产品和服务，人工智能正加速向我们走来。 大模型赋能千行百业 随着ChatGPT、DeepSeek等大模型横空出世，生成式人工智能迎来爆发式增长，成为驱动经济社会发展 的强大引擎。 人工智能发展有多快？科大讯飞研究院副院长殷兵告诉记者，过去一年，随着国内"百模大战"逐渐收 官，业界巨头下场、头部玩家凸显、新锐势力脱颖而出，人工智能成为全球科技博弈最紧张、最激烈的 领域，大模型依然是其中的主角。 机器人加快走进现实 作为通用人工智能的理想载体，以人形机器人为代表的具身智能发展是近年来人工智能领域的焦点。 搭载高度仿生的躯干构型和拟人化的运动控制能力，全身多达20个自由度，能以每小时10公里的速度稳 定奔跑，还具备在复杂泛化地形平稳移动、抗冲击干扰等运动功能……它就是优必选科技联合北京人形 机器人创新中心推出的全尺寸科研教育人形机器人——天工行者。 在技术研发方面，高校可利用天工行者开展 ...

Core Viewpoint - The launch of the 300 billion parameter Kunlun model by China National Petroleum Corporation, China Mobile, Huawei, and iFlytek marks a significant milestone in the application of artificial intelligence within the energy and chemical industry, establishing a benchmark for state-owned enterprises in leveraging AI technology for business scenarios [1][7]. Group 1: Model Development and Capabilities - The Kunlun model is the first industry model to be officially registered and evaluated by the China Academy of Information and Communications Technology, showcasing a comprehensive upgrade in capabilities, with language model parameters increasing from 70 billion to 300 billion, visual model parameters from 300 million to 4.4 billion, and multi-modal model parameters from 16 billion to 80 billion [2]. - The model supports 100 innovative application scenarios across the entire industry chain, including exploration, refining, equipment manufacturing, and sales, addressing key areas such as carbonate rock slip fracture identification [2][3]. Group 2: iFlytek's Role and Strategy - iFlytek has played a crucial role in the development of the Kunlun model, providing significant technical support and creating 32 application scenarios, including digital document management and intelligent analysis for the natural gas market [3][4]. - The company emphasizes a "1+N" strategy, focusing on a leading foundational model combined with vertical applications across various industries, thereby enhancing the high-quality development of enterprises [4][6]. Group 3: Market Position and Future Outlook - iFlytek's self-developed models, such as the Xinghuo model, have surpassed ChatGPT in several dimensions, including text generation and knowledge Q&A, establishing a strong position in the domestic market [4][6]. - The company has supported numerous state-owned enterprises in building representative industry models, with its multilingual technology covering 85 languages and reaching over 200 countries and regions [6]. - According to "Intelligent Super Parameters," iFlytek leads in both the number and value of bids for large model-related procurement projects initiated by state-owned enterprises in 2024 [6].

金十数据整理：每日科技要闻速递（6月18日） 1. 微信小范围测试聊天记录备份至外部存储设备。 2. 科大讯飞：讯飞星火X1升级版预计于今年7月推出。 3. 美团CEO王兴再次减持理想汽车，4天套现6亿港元。 4. AI初创公司Anysphere获风投青睐 估值超180亿美元. 5. 马斯克的XAI正在洽谈通过股权融资筹集43亿美元。 6. 腾讯元宝上线AI编程模式：双栏界面实现边提需求边写代码。 7. 我国载人登月又有新突破，梦舟飞船零高度逃逸飞行试验成功。 8. 消息称微软正开发下一代Xbox Cloud Gaming云游戏服务器。 9. 据CFM闪存市场，DDR4内存条报价持续上调现货市场成交已现乏力。 10. 宝能汽车回应"解散、清算"报道：公司正常经营，还有新车即将上市。 11. 我国磁悬浮技术获新突破，成功在1000米距离内将1.1吨重试验车加速至650公里/小时。 12. 野村证券预测Meta最早第四季度推出下一代ASIC芯片MTIA T-V1：规格有望超越英伟达"Rubin"。 ...

Core Insights - The chairman of iFlytek, Liu Qingfeng, emphasized the challenges of developing algorithms based on domestic computing power compared to using overseas solutions, noting a potential delay of three months in algorithm development [1] - iFlytek is committed to a self-controlled path in algorithm development, citing the need for independence to avoid being constrained by external factors [1] - The company has seen significant success in various sectors, including education and legal, with its AI models outperforming competitors in specific applications [3][4] Group 1: Business Development - iFlytek's business is thriving across C-end, B-end, and G-end sectors, with numerous benchmark cases emerging from state-owned enterprises [3] - The company has achieved comprehensive leadership in the financial and state-owned enterprise markets, focusing on solidifying its foundational models and standard products [4] - The star product, the X1 model, has shown impressive performance in medical and legal applications, with accuracy rates significantly surpassing competitors [3][4] Group 2: Consumer Market Performance - iFlytek's consumer products, including learning machines and translation devices, have dominated sales on major platforms, with a 25% year-on-year revenue growth [6] - The total user base for the SaaS product "Tingjian" has reached 90 million, reflecting a 54% year-on-year increase [6] - The overseas market has seen explosive growth, with the number of overseas developers increasing by 31.9% and AI marketing revenue growing 25 times year-on-year [6] Group 3: Future Outlook - Liu Qingfeng believes that a nation's global influence will be measured by the number of sovereign large models it can provide, positioning iFlytek to contribute significantly to this landscape [6] - The year 2025 is anticipated to mark the beginning of widespread AI adoption, with projections indicating that 17.7% of China's population will be users of generative AI by the end of 2024 [6] - iFlytek aims to capture the foundational benefits of the AI era by focusing on self-control, integration of software and hardware, and addressing industry-specific needs [6]

2025年高考大幕虽已落下，但关于数学科目难度的讨论热度不减。 《每日经济新闻》记者（以下简称"每经记者"）选取今年的全国新课标数学I卷作为考题，对DeepSeek-R1、腾讯混元T1、OpenAI的o3、谷歌的Gemini 2.5 Pro和xAI的Grok3等十款AI推理大模型进行了测评，以检验当今主流AI推理大模型的数学能力。 测评结果显示，国产大模型DeepSeek-R1与腾讯混元T1以零错误并列榜首。而被马斯克称为"地表最强AI"的Grok 3却遭遇"滑铁卢"，排名倒数第三。 本次测评以2025年全国新课标数学I卷（总分150分）作为考题。但每经记者在测试中发现，部分AI推理模型以"重要考试期间"为由拒绝对包含试题的图片进 行识别和解答。 为了让所有参评大模型站在同一起跑线，测评移除了试卷中所有需要分析图形和图表的题目，形成一份有效总分为117分的标准化试卷。 同时，对于谷歌Gemini 2.5 Pro等没有这一限制的推理模型，仍将以150分的完整试卷进行测试，旨在测试推理大模型所能达到的最高水平。 扣分标准上，每经记者在选择题和填空题上都遵循了高考评卷的扣分标准，但对于解答题，本次测评只根据结果计算 ...

Core Insights - The chairman of iFlytek, Liu Qingfeng, emphasized the concept of "root dividend," focusing on self-control, integration of software and hardware, and deep industry engagement to drive innovation in large model technology and achieve true industrial advantages [1][2] - iFlytek's diverse applications showcase its advancements, including the X1 medical model reaching the level of chief physicians in top hospitals and the X1 legal model excelling in legal recommendations [1][3] - The company reported significant growth across its business segments, with a 25% year-on-year increase in revenue from products like learning machines and translation devices, and a 54% increase in users for its "Tingjian SaaS" software [1][3] B-end Developments - iFlytek has secured leading clients in the B-end sector, collaborating with major companies like PetroChina to develop the Kunlun large model and working with the State Energy Group on an intelligent evaluation assistant [2] G-end Applications - The AI-assisted case handling system for procuratorates has been launched for trial use, indicating progress in G-end applications [3] - The overseas market has shown rapid growth, with a 25-fold increase in AI marketing revenue and a threefold increase in C-end hardware sales from January to May [3] Technological Innovations - Liu Qingfeng highlighted three major technological innovations in iFlytek's large models, including a unified model training and reasoning method, reinforcement learning with fine-grained feedback, and a self-controlled reinforcement learning algorithm package [3] - The company is committed to training models based on domestic computing power, facing higher technical barriers and longer R&D cycles compared to overseas computing power, which is seen as essential for true self-control [3] Strategic Vision - Over its 26 years, iFlytek has transitioned from speech synthesis to cognitive intelligence and large models, with Liu Qingfeng stating the company aims to lead in foundational model technology while addressing real industry needs to enhance productivity and creativity [3]

高考后多家媒体针对语数外三科的深度推理模型答题横评中，科大讯飞旗下讯飞星火X1深度推理大模型，更是超 越一众主流模型，实现 综合 答题效果第一，以绝对优势稳居大模型高考答题的第一梯队，成为 2025 年最会高考 同时也是最懂高考的国产大模型。 底座模型技术领先，只是科大讯飞 AI 学习机能够精准押中高考原题的原因之一。在 AI 赋能教育的过程中，技术 只是影响因素之一，但更核心的仍然是对教育的理解和积累。大模型本身的能力、学习资源的多少，只是衡量学 习机好坏的标准之一，而非决定因素。 2025年高考终于落下帷幕。和往年相比，AI在高考中的浓度越来越高。AI对于高考的介入，不只是高考结束后各 家AI大模型在答题中的比拼，更是深入到高考从考场、阅卷、打分乃至学生的日常学习之中。 今年4月份，教育部等九部门联合印发《关于加快推进教育数字化的意见》，强调全面推进智能化，促进 人工智 能 助力教育变革。意见中特别提到，要推动课程、教材、教学数字化变革，核心目的就是利用人工智能来赋能教 育，尤其是"通过智能学伴、数字导师等探索人机协同教学新模式，实现人工智能驱动的大规模因材施教，提高教 育教学效率和质量。" 这正是近年来 ...

小米中国区近日迎来人事调整，小米集团副总裁、中国区总裁、新零售部总经理王晓雁兼任新零售部小米 之家总经理。据悉，王晓雁兼任小米之家总经理之后，原小米之家总经理王辉将转任销售管理部总经理， 继续向王晓雁汇报。 要闻提示 NEWS REMIND 1.小米中国区人事调整：副总裁王晓雁兼任小米之家总经理 2.DeepSeek核心高管离职创业：瞄准Agent赛道 3.传大疆影像系统奠基人，团队负责人离职 4.损失达几十亿？美的回应北美空调事件：不存在缺陷，系主动召回 5.比亚迪首次杀入日本进口车前十！今年底要开100家门店 6.特斯拉前四月 全球销量榜To p10：九国"跌跌不休"，只有加拿大涨了 7. 苹果 WWDC：AI加持siri尚需时日，向开发者开放基础模型，界面设计液态玻璃首秀 8.Neuralink和Grok合作，脑机芯片让渐冻症患者重新"说话" 今日头条 HEADLINE NEWS 小米中国区人事调整：副总裁王晓雁兼任小米之家总经理 2024年底，小米集团副总裁、中国区新零售部总经理马骥离职，王晓雁开始兼任新零售部总经理一职，如 今再兼任小米之家总经理，可谓身兼多职。王晓雁于2019年加入小米，先后担任中 ...

6月8日下午，2025年高考英语科目考试结束。在今年高考英语北京卷的作文题目中，李华又一次成为"主角"。 与此同时，这场"语言能力大考"也成为AI大模型的试金石——从精准审题到地道表达，从逻辑连贯到文化适配，大语言模型能否作为"李华"完成高考英语作 文题目的独特要求？在各个大模型均已具备"深度思考"模式的当下，"AI考生"能拿到多少分？对此，新京报贝壳财经对国内外主流大语言模型进行了测评。 本次测评选择2025年高考英语北京卷作文题目，测评对象为DeepSeek R1、ChatGPT o3、通义千问Qwen3、腾讯混元T1、讯飞星火X1、百度文心X1六款深度 推理模型产品，并邀请北京市十一学校一分校英语老师韩宪昌、深圳中学英语教师赵文嘉参照往年高考评分标准对上述大模型生成的英语作文打分并作点 评。 本着客观公正和"盲评"原则，韩宪昌老师和赵文嘉老师给出了专业评分。本次参赛的六名大模型"AI考生"得分揭晓后由高到低的排名分别为：讯飞星火 X1、DeepSeek R1、百度文心X1、通义千问Qwen3、ChatGPT o3、腾讯混元T1。 具体来看，2025高考英语北京卷作文题目的要求为： 假设你是红星中学高三学 ...