A股科技赛道午后延续强势，板块题材上，CPO、元件、商业航天板块活跃，煤炭板块调整，人工智能 AIETF（515070）午后持续攀升，盘中涨幅扩大至3.61%，其持仓股新易盛大涨8.43%，中际旭创涨 7.29%，寒武纪-U、光迅科技、复旦微电等股领涨。 消息面上，谷歌在 Gemini 应用中正式推出了 Gemini 3 Deep Think 模式，面向 Google AI Ultra 订阅用 户。这一新模式显著提升了推理能力，旨在应对复杂的数学、科学和逻辑问题。Gemini3Deep Think 在 多个严格的基准测试中表现出色，比如在 "人类最后的考试" 中，未使用工具的情况下取得了41.0% 的 成绩，而在 ARC-AGI-2测试中，使用代码执行时更是达到了前所未有的45.1%。这种出色表现得益于其 采用了先进的并行推理技术，能够同时探索多个假设。 （文章来源：每日经济新闻） 人工智能AIETF（515070）跟踪CS人工智能主题指数（930713），成分股选取为人工智能提供技术、 基础资源以及应用端个股，聚集人工智能产业链上中游，俗称"机器人"大脑"缔造者"，万物互联"地 基"。前十大权重股包括中际旭 ...

太劲爆了！ 不过半月，谷歌DeepMind终于放出了IMO最强金牌模型——Gemini 3 Deep Think。 这一次，谷歌为其注入了全新的血液——Gemini 3。 凭借着「并行思考」能力，Gemini 3 Deep Think可以搞定超高难度的数学、科学难题！ 在基准测试中，Deep Think全面碾压Gemini 3 Pro，尤其是在HLE上，未用工具拿下了41%高分。 同时在ARC-AGI-2上，以45.1%成绩领跑全球。 下面实例中，同一个指令，让Gemini 3 Pro和Deep Think版基于一张博物馆展馆屋顶的草图，创建一个精确的交互式3D场景。 显然，后者在还原度上，与原图几乎是1:1复刻，并在交互上，光影变化符合物理逻辑。 今年夏天，Gemini 2.5 Deep Think分别在IMO、ICPC国际大赛中，拿下了金牌的战绩。 今天，Gemini 3 Deep Think已在Gemini App上线，所有Ultra用户即可体验。 最强IMO金牌模型来了 Gemini 3 Deep Think正式开启了「深度思考」新纪元，让智能的边界再次拓展。 Gemini 3 Deep Think基 ...

作者 ｜ 木子、高允毅 刚刚， Gemini 3 的 Deep Think 模式 终于 正式上线 了。 顾名思义，这是 Gemini 3 的深度思考模式， 推理能力显著加强 ， 能处理复杂、多步骤，以及更多 创新的问题， 还可以搞定超难的科学问题和数学题！ ARC-AGI-2 则将任务升级为多步骤、递归、隐藏规则等，是 更接近"类人智慧"的高阶推理场景 。 其中，Gemini 3 Deep Think 正确率达 45.1% ，比非深度思考模式的 Gemini 3 Pro（正确率 31.1%）高出了 14%。而在这项测试中，GPT-5 Pro 的正确率仅有 18.3% 。 是 ARC-AGI、HLE 等 多项权威测评中的第一名 先来看看 Gemini 3 Deep Think 是怎么一回事。 在公认的大模型最难测试之一、全球 最接近"通用智能（AGI）核心能力"验证 的基准测试 ARC-AGI 中，Gemini 3 Deep Think 在 2 个榜单中均 拔得头筹 。 其中， ARC-AGI-1 主要测 模型的基础抽象推理 。在这项测试中，Gemini 3 Deep Think 的答题正确 率排第一，达到了 ...

Jay 发自 凹非寺 量子位 | 公众号 QbitAI 奥特曼又得拉响红色警报了。 刚刚，谷歌再次扔出重磅炸弹—— Gemini 3 Deep Think 正式上线！ 这款谷歌最新最强模型，推理能力确实有点离谱。 轻松把草图变成逼真3D场景，不仅结构还原到位，就连镂空花纹与光影都处理得明明白白。 甚至有网友拿它搞起了视觉艺术，一人一AI在虚拟宇宙里「不知天地为何物」。 看完这些demo，估计奥特曼只得再次咬牙切齿送上「happy for u」了。 （doge） 几句话就能搭出个3D多米诺骨牌解压游戏，运行相当丝滑。 Ultra用户今天就能通过Gemini聊天框里的「Deep Think」选项用起来了～ 高歌猛进的Gemini，又一次屠榜 不给对手任何喘息的机会，Gemini 3 Pro刚给OpenAI按在地上锤完，谷歌转手又扔出一重磅炸弹——Gemini 3 Deep Think。 相比之前的模型，新版本在复杂数学、科学推理和逻辑问题上都有大幅提升，旨在攻克那些连最强模型都难以解决的数学、科学和逻辑问题。 具体来说，在「深度思考」模式下，Gemini会开启迭代推理，能多轮打磨代码，生成更精细的程序，从而在可视 ...

Gemini 3 Deep Think is now available for Google AI Ultra subscribers in the @GeminiApp, incorporating our gold medal winning IMO and ICPC technologies! 🏅With its parallel thinking capabilities it can tackle highly complex maths & science problems - enjoy! https://t.co/5BXWmYyor6 ...

Investment Rating - The report maintains a "Recommendation" rating for the computer industry, expecting the industry index to rise more than 5% over the next 3-6 months compared to the benchmark index [4][35]. Core Insights - Gemini 3, launched by Google, significantly enhances its AI competition position with superior reasoning and multimodal capabilities, achieving a groundbreaking Elo score of 1501 on the LMArena leaderboard [10][11]. - The model excels in various benchmarks, including a 91.9% accuracy in the GPQA Diamond test and a 23.4% score in the MathArena Apex test, showcasing its advanced performance in reasoning and mathematics [10][11]. - Google has introduced the Antigravity platform, which integrates Gemini 3's capabilities, allowing for autonomous planning and execution of complex software engineering tasks [17][20]. - The "chip-model-ecosystem" strategy positions Google with a competitive edge, leveraging self-developed Trillium TPU chips that enhance computing power by 4 times and reduce energy consumption by 67% [20][21]. Summary by Sections Section 1: Gemini 3 Performance Evolution - Gemini 3 demonstrates significant technical breakthroughs, outperforming its predecessor in all key AI benchmarks, including achieving a 37.5% score in the "Humanity's Last Exam" without tools [10][11]. - The model's multimodal understanding is highlighted by its performance in MMMU-Pro and Video-MMMU tests, scoring 81% and 87.6% respectively [10][11]. Section 2: Gemini 3 Deep Think - The Deep Think mode of Gemini 3 expands its capabilities, achieving a 41.0% score in the "Humanity's Last Exam" without tools and a 93.8% accuracy in the GPQA Diamond test [15][16]. Section 3: Antigravity Platform - Antigravity enhances the development experience by providing a dedicated interface for agents, allowing them to operate editors, terminals, and browsers autonomously [17][20]. Section 4: Chip-Model-Ecosystem Strategy - Google's strategy integrates hardware, models, and ecosystem, achieving significant results with Gemini series models, which have seen over 6.5 billion monthly active users and a 70% adoption rate among cloud customers [20][21]. Section 5: Investment Recommendations - The report suggests focusing on specific sectors within AI, including domestic computing power and enterprise services, highlighting companies like Cambricon, Alibaba, and Kingsoft Office [22][24].

Core Insights - Gemini 3 has been launched, showcasing superior capabilities compared to its predecessor Gemini 2.5 Pro and competitors like Claude Sonnet 4.5 and GPT-5.1 [1][3] - The model can generate 3D models, websites, and even open-world games with a single sentence, indicating a significant advancement in AI capabilities [2][5] Performance Metrics - Gemini 3 Pro outperformed Gemini 2.5 Pro in various benchmarks, achieving notable scores such as 37.5% in "Humanity's Last Exam" and 31.1% in "ARC-AGI-2" [3][7][8] - In mathematical problem-solving, Gemini 3 Pro scored 95.0% in "AIME 2025" and 23.4% in "MathArena Apex," significantly surpassing competitors [4][7] - The model achieved a score of 100% in code execution tasks, demonstrating its advanced coding capabilities [3] New Features and Applications - Gemini 3 introduces a "Generative UI" feature, allowing users to interact with AI-generated content in a more dynamic way, moving beyond traditional Q&A formats [15][39] - The model can assist in practical tasks, such as booking rentals or planning trips, by integrating data from various Google services [5][7] - Gemini 3 Pro can create high-quality interactive SVGs and even complete websites in a matter of minutes, showcasing its potential to disrupt traditional design and development roles [9][26][39] Industry Impact - The advancements in Gemini 3 may pose a threat to designers and programmers, as the model can perform tasks that typically require human expertise [9][19] - The introduction of tools like Antigravity positions Gemini 3 as a productivity assistant for developers, potentially lowering the technical barriers in coding [38][39]

Core Insights - Google has launched its new flagship AI model, Gemini 3 Pro, which is touted as the "strongest reasoning + multimodal + ambient programming" AI to date, outperforming competitors like OpenAI's GPT-5.1 in benchmark tests [1][3][9] Performance Highlights - Gemini 3 Pro achieved significant improvements over its predecessor, Gemini 2.5 Pro, and outperformed GPT-5.1 in various benchmarks, including: - Humanity's Last Exam (HLE): 45.8% (highest score) without tools [4][5] - GPQA Diamond: 91.9% [4][17] - AIME 2025 (Mathematics): 95.0% [4][18] - Vending-Bench 2: $5,478.16 in net worth [4][18] Multimodal Capabilities - The model excels in multimodal understanding, scoring 81.0% in MMMU-Pro and 87.6% in Video-MMMU, showcasing its ability to process and reason across different types of data [19][22] - Gemini 3 can interpret complex scientific concepts and generate high-fidelity visual code, enhancing its utility in various fields [22][24] Ambient Programming - Gemini 3 Pro has advanced ambient programming capabilities, allowing developers to create interactive applications with simple prompts, significantly improving the development process [14][31] - The model scored 1487 Elo in the WebDev Arena, indicating its strong performance in web development tasks [31][32] Deep Think Mode - The introduction of Gemini 3 Deep Think mode marks a new era in AI, achieving exceptional results in challenging benchmarks, including 41% in HLE and 93.8% in GPQA Diamond [25][28] - This mode enhances the model's ability to tackle complex problems and demonstrates its potential for advanced reasoning [25][28] Developer Integration - Gemini 3 is integrated into various platforms, including Google AI Studio and Google Antigravity, allowing developers to leverage its capabilities for building sophisticated applications [36][42] - The model's training was completed on Google's TPU, reinforcing its competitive edge in the AI landscape [54]

Model Performance & Benchmarks - Gemini 3 surpasses previous Frontier models in benchmarks, demonstrating significant advancements in AI capabilities [1] - Gemini 3 achieves 458% with code execution and search on Humanity's last exam, compared to Gemini 25% Pro at 21%, Cloud Sonnet 45% at 13%, and GBT 51% at 265% [2] - On the Vending Bench benchmark, Gemini 3's net worth reached $547816%, significantly outperforming Cloud Sonnet 45% at $3800 [4] - Gemini 3 Deep Think scores 41% on Humanity's Last Exam, compared to Gemini 3 Pro at 375%, Claude Sonnet 45% at 13%, GPT5 Pro at 30%, and GPT 51% at 265% [9][10] - Gemini 3 Deepthink achieves 451% on Arc AGI2 visual reasoning puzzles, a 10x improvement over Gemini 25% Pro [12] Enterprise Applications & Features - Boxcom's benchmark shows a 22-point performance increase for Gemini 3 Pro versus Gemini 25% Pro, with scores of 85% and 63% respectively [6] - Industry subsets in Boxcom's benchmark show significant performance jumps: Healthcare and Life Sciences (45% to 94%), Media and Entertainment (47% to 92%), and Financial Services (51% to 60%) [6] - Gemini 3 excels in complex multi-step reasoning and task automation, as highlighted by Box's new benchmark [7] - Gemini 3 supports multiple modalities, including text, images, video, audio, and code, with a unique focus on video understanding [12] - Gemini 3 can analyze YouTube videos frame by frame, understanding the content in detail [13] Google Integration & New Products - Gemini 3 is integrated into Google Search, dynamically generating user interfaces based on user queries [17] - Google launched anti-gravity, a VS Code fork coding platform that supports Gemini models and other models like GPTOSS and Anthropic's Sonnet [20] - The updated Gemini app features Gemini Agent capability, enabling the AI to complete real tasks on the user's behalf and create dynamic UIs [24] Model Architecture & Specifications - Gemini 3 is a brand new foundation model, not a modification of a prior model [27] - The model accepts text, images, audio, and video files as inputs, with a token context window of up to 1 million and output tokens of 64000 [28] - Gemini 3 is a sparse mixture of experts model built on Google's custom TPU architecture for both pre-training and inference [28]

AI Model Advancement - Gemini 3 Deep Think is introduced as a state-of-the-art (SOTA) model, surpassing Gemini 3 Pro [1]