刚测试了下，OpenAI的GPT OSS 120B在Apple M3 Ultra 512G本地运行大概是43-45t/s，属于相当不错且可用的速度。并且让GPT-4.5出题来测试性能，GPT-4.5评估GPT OSS回答后，给出了相当不错的评价。 https://t.co/b6shhCm0mGSam Altman (@sama):gpt-oss is out!we made an open model that performs at the level of o4-mini and runs on a high-end laptop (WTF!!)(and a smaller one that runs on a phone).super proud of the team; big triumph of technology. ...

（注：我们会对线索进行核实。您的隐私将严格保密。） OpenAI 发 布 两 款 " 开 源 " 和 免 费 使 用 的 AI 模 型 ， GPT-oss-120b 和 GPT-oss-20b 。 这 次 发 布 是 OpenAI 自发布GPT-2以来，首次推出新的"开源"大语言模型。 OpenAI CEO 山姆·奥尔特曼在社交媒体表示：" GPT -oss是一个重大突破，这是最先进的开放权重 推理模型，具有与o4-mini相当的强大现实世界性能，可以在你自己的电脑（或手机的较小版本）上 本地运行。"他透露公司将在未来几天里带来许多新东西。 微信编辑 | 七三 第 一 财 经 持 续 追 踪 财 经 热 点 。 若 您 掌 握 公 司 动 态 、 行 业 趋 势 、 金 融 事 件 等 有 价 值 的 线 索 ， 欢 迎 提 供 。 专 用 邮 箱 ： bianjibu@yicai.com 2025.08. 06 本文字数：304，阅读时长大约1分钟 作者 | 一财科技 推荐阅读 特朗普称已收到访华邀请，外交部回应 ...

Group 1 - Hema X member stores will all cease operations as part of a business adjustment, with the last store closing on August 31 [3] - Li Xiang, CEO of Li Auto, announced a unified configuration for the Li i8 model, aiming for better sales performance in the 300,000 to 400,000 yuan price range [4] - Goldman Sachs is reportedly planning to acquire a stake in Froneri, the world's second-largest ice cream manufacturer, for 15 billion euros (approximately 125 billion yuan) [6] Group 2 - Geely Auto is undergoing significant adjustments in its intelligent driving team, with plans still under discussion [7] - Haidilao denied rumors of transitioning to a semi-self-service model, calling the claims false [7] - NetEase experienced a major outage affecting multiple games, attributed to internal server issues, with downtime exceeding two hours [9] Group 3 - JD.com is set to launch its first large discount supermarket format in August, with stores in Jiangsu and Hebei, featuring a wide range of products at lower prices [9] - Chery has mandated a 30% reduction in meetings to improve efficiency, reflecting on past inefficiencies [10] - Alibaba is launching a new membership system that integrates various services, with over 6% of positions focused on AI in its 2026 campus recruitment [14] Group 4 - Ant Group sold its remaining shares in Indian digital payment company Paytm for approximately $454 million, marking a complete exit from the company [14] - Tata Motors appointed its CFO as the new CEO of Jaguar Land Rover, replacing the previous CEO after three years [13] - Tencent led an investment round in the Uzbek fintech company Uzum, valuing it at around $1.5 billion [16] Group 5 - OpenAI released two open-weight AI models capable of mimicking human reasoning processes, available on the Hugging Face platform [20] - The Gates Foundation announced a $2.5 billion investment in women's health research, focusing on five key areas [20] - The global tablet shipment volume increased by 9% year-on-year in Q2 2025, with Apple maintaining the leading position [24]

Group 1 - Former Google executive Mo Gawdat predicts that AI will eliminate the middle class within the next decade, stating that most knowledge workers, including programmers and CEOs, will be replaced by AI, leaving only the top 0.1% of earners [2] - Alibaba's Tongyi Qianwen has launched Qwen-Image, a new 20 billion parameter MMDiT model for image generation, achieving significant advancements in complex text rendering and precise image editing, and demonstrating state-of-the-art performance in various generation and editing tasks [2] - Tencent's AI workspace product ima has introduced new features that allow users to upload files to generate AI podcasts and support for folder imports and knowledge management, enhancing knowledge sharing capabilities [2] Group 2 - OpenAI announced that ChatGPT's weekly active users will reach 700 million, a fourfold increase compared to the previous year, with daily message volume exceeding 3 billion, indicating accelerated growth in user engagement [2]

gpt-oss-20b可在16GB内存的设备上运行，gpt-oss-120b需要约80GB内存，适合包括Mac电脑在内的个人 设备。 gpt-oss-120b在核心推理基准测试中接近o4-mini，gpt-oss-20b则达到或超过o3-mini，且在特定任务上表 现更优。 这些模型设计为低成本选项，支持本地运行、工具使用和思维链处理，适合开发者和研究人员定制。 OpenAI发布了gpt-oss-120b和gpt-oss-20b两款开放权重语言模型，这是自2019年GPT-2以来首次。 ...

当天早些时候，OpenAI发布了两个开放权重AI模型，分别是GPT-oss-120b和GPT-oss-20b。 微软首席执行官萨蒂亚·纳德拉当地时间8月5日发文称，很高兴将OpenAI的gpt-oss模型引入Azure"AI应 用和智能体工厂"AI Foundry，并通过Foundry Local将其带到Windows平台。纳德拉称，这是混合AI的实 际应用：用户可以灵活组合不同模型，优化性能与成本，并直接在数据所在的位置进行处理。 ...

RT Taelin (@VictorTaelin)My initial impression on OpenAI's OSS model is aligned with what they advertised. It does feel closer to o3 than to other open models, except it is much faster and cheaper. Some providers offer it at 3000 tokens/s, which is insane. It is definitely smarter than Kimi K2, R1 and Qwen 3. I tested all models for a bit, and got very decisive results in favor of OpenAI-OSS-120b.Unfortunately, there is one thing these models can't do yet - my damn job. So, hope you guys have fun. I'll be b ...

Financial Support for New Industrialization - The People's Bank of China and other regulatory bodies issued guidelines to support new industrialization, focusing on key sectors like integrated circuits and industrial mother machines [1][2] - Banks are encouraged to provide long-term financing for technology breakthroughs and facilitate easier access to capital for companies achieving core technology advancements [1][2] Emerging Industries and Financing - New industries such as information technology, renewable energy, and biomedicine will have access to multi-tiered capital markets for financing [2] - Long-term funds from government investment funds and insurance will focus on future manufacturing and energy sectors [2] Support for Small and Medium Enterprises - Financial institutions are urged to reduce reliance on guarantees and provide financing based on data and asset credit [2] - A national credit information platform for small and micro enterprises is being developed to facilitate easier access to credit [2] Green Transition Financing - Financial support will be directed towards high-carbon industries that meet green transformation criteria, with a focus on green credit and bonds [2] - A specialized financial standard system will be established to enhance funding for green projects [2] Digital Integration and Services - Digital infrastructure projects like 5G and industrial internet will receive long-term loans and financing options [2] - Banks are developing digital platforms to provide one-stop services for financing and settlement, improving efficiency for small businesses [2] Risk Management in Financial Institutions - Financial institutions are required to monitor the use of funds to prevent misuse and ensure compliance with regulations [3] - Joint risk assessments will be conducted to share high-risk information and manage potential financial risks [3] Market Trends and Predictions - Major financial institutions have warned clients to prepare for potential declines in U.S. stock prices, with predictions of a 10% to 15% correction in the S&P 500 index [17][18] - The retail forecast for passenger vehicles in 2025 has been slightly adjusted upward, indicating a growth of 6% [19]

Core Insights - The article provides a comprehensive overview of large language models (LLMs), their definitions, architectures, capabilities, and notable developments in the field [3][6][12]. Group 1: Definition and Characteristics of LLMs - Large Language Models (LLMs) are deep learning models trained on vast amounts of text data, capable of understanding and generating natural language [3][6]. - Key features of modern LLMs include large-scale parameters (e.g., GPT-3 with 175 billion parameters), Transformer architecture, pre-training followed by fine-tuning, and multi-task adaptability [6][12]. Group 2: LLM Development and Architecture - The Transformer architecture, introduced by Google in 2017, is the foundational technology for LLMs, consisting of an encoder and decoder [9]. - Encoder-only architectures, like BERT, excel in text understanding tasks, while decoder-only architectures, such as GPT, are optimized for text generation [10][11]. Group 3: Core Capabilities of LLMs - LLMs can generate coherent text, assist in coding, answer factual questions, and perform multi-step reasoning [12][13]. - They also excel in text understanding and conversion tasks, such as summarization and sentiment analysis [13]. Group 4: Notable LLMs and Their Features - The GPT series by OpenAI is a key player in LLM development, known for its strong general capabilities and continuous innovation [15][16]. - Meta's Llama series emphasizes open-source development and multi-modal capabilities, significantly impacting the AI community [17][18]. - Alibaba's Qwen series focuses on comprehensive open-source models with strong support for Chinese and multi-language tasks [18]. Group 5: Visual Foundation Models - Visual Foundation Models are essential for processing visual inputs, enabling the connection between visual data and LLMs [25]. - They utilize architectures like Vision Transformers (ViT) and hybrid models combining CNNs and Transformers for various tasks, including image classification and cross-modal understanding [26][27]. Group 6: Speech Large Models - Speech large models are designed to handle various speech-related tasks, leveraging large-scale speech data for training [31]. - They primarily use Transformer architectures to capture long-range dependencies in speech data, facilitating tasks like speech recognition and translation [32][36]. Group 7: Multi-Modal Large Models (MLLMs) - Multi-modal large models can process and understand multiple types of data, such as text, images, and audio, enabling complex interactions [39]. - Their architecture typically includes pre-trained modal encoders, a large language model, and a modal decoder for generating outputs [40]. Group 8: Reasoning Large Models - Reasoning large models enhance the reasoning capabilities of LLMs through optimized prompting and external knowledge integration [43][44]. - They focus on improving the accuracy and controllability of complex tasks without fundamentally altering the model structure [45].

Model Release & Performance - Anthropic 发布了 Claude Opus 4.1%，是对 Claude Opus 4 的升级，尤其在 Agentic 任务、真实世界编码和推理方面 [1] - SWEBench verified 基准测试中，Claude Opus 4.1% 的得分从 Opus 4 的 72.5% 提升至 74.5%，提升了 2 个百分点 [3] - Terminal Bench 基准测试中，Claude Opus 4.1% 的终端使用能力从 39.2% 提升至 43.3%，提升了 4.1 个百分点 [4] - GPQA Diamond（研究生水平推理）基准测试中，Claude Opus 4.1% 的得分从 79.6% 提升至 80.9%，提升了 1.3 个百分点 [4] - Towbench（Agentic 工具使用）基准测试中，Claude Opus 4.1% 在零售方面的得分从 81.4% 提升至 82.4%，提升了 1 个百分点，但在航空方面从 59.6% 下降至 56%，下降了 3.6 个百分点 [5] - 多语言问答基准测试中，Claude Opus 4.1% 的得分从 88.8% 提升至 89.5%，提升了 0.7 个百分点 [5] - Amy 2025 基准测试中，Claude Opus 4.1% 的得分提升了 2.5 个百分点至 78% [5] Competitive Positioning & Future Outlook - 在 SWEBench 和 Terminal Bench 基准测试中，Claude Opus 4.1% 优于 OpenAI 的 GPT-3 和 Gemini 1.5 Pro [5] - 在 GPQA Diamond 和 Agentic 工具使用基准测试中，Claude Opus 4.1% 不及 OpenAI 的 GPT-3 和 Gemini 1.5 Pro [6] - 在高中数学竞赛基准测试中，Claude Opus 4.1% 的得分低于 OpenAI 的 GPT-3 (88.9%) 和 Gemini 1.5 Pro (88%)，仅为 78% [6] - Claude 目前被广泛认为是市场上最佳的编码模型，尤其擅长 Agentic 编码和 Agent-driven 开发 [7]