Workflow
开源AI
icon
Search documents
智谱股价飙升24%,外国网友直呼“GLM-5是最好的开源模型”
Ge Long Hui· 2026-02-24 05:14
智谱称,GLM-5是一款旨在推动编程范式从"Vibe Coding"(氛围编程)转向"Agentic Engineering"(智能体工程)的下一代基础模型。GLM-5在前代模型GLM-4.5 的智能体、推理与编程(Agentic, Reasoning and Coding, ARC)能力基础上,采用稀疏注意力(DeepSeek Sparse Attention,DSA)以大幅降低推理成本,同时保持 长上下文能力无损。为了让模型更好地与各类任务对齐,GLM-5构建了一套新型异步强化学习(RL)基础设施,通过将生成过程与训练过程解耦,从而大幅提 升了后训练的迭代效率。此外,GLM-5还提出了全新的异步Agent强化学习算法,进一步提升强化学习的效果,使模型能够更有效地从复杂、长程交互中学 习。基于上述创新,GLM-5在主流的开放基准测试中实现了SOTA性能。最关键的是,GLM-5在真实世界编程任务中展现出前所未有的能力,在处理端到端 软件工程挑战方面超越了此前所有开源基线。 智谱(2513.HK)今日重拾升势,盘中股价一度飙升24.64%,报698港元。 智谱于2月12日正式推出新一代旗舰模型GLM-5,其在编 ...
早报|美团2025年预亏超233亿元;美军将向中东增派第二航母;携程、高德等6家出行平台被约谈;荣耀原研发部总裁被批准逮捕
虎嗅APP· 2026-02-14 00:28
【SpaceX据称拟在IPO中采用双重股权结构以强化马斯克控制权】 财联社2月14日电,据知情人士透露,SpaceX正考虑在今年计划中的首次公开募股中采用双重股权结构,此举 与其亿万富翁创始人埃隆·马斯克为特斯拉公司提出的策略如出一辙。 昨夜今晨 【特朗普证实美军将向中东地区派出第二艘航母】 财联社2月14日电,美国总统特朗普当地时间2月13日证实,美军将向中东地区派出第二个航空母舰打击群,以 此施压伊朗同美国达成协议。 特朗普在白宫接受媒体采访时说,如果美国与伊朗达不成协议,"我们就需要它(第二艘航母)"。他同时表 示,如果达成协议,美军航母会"很快离开"。 大家早上好!这里是今天的早报,每天早上,我都会在这里跟你聊聊昨夜今晨发生了哪些大事儿。 特朗普表示,美军已经在中东部署一个航母打击群,如果需要,增加部署就会到位。他再次威胁伊朗,如果谈 判不成功,结果会"糟糕"。 【 王毅在慕尼黑会见美国国务卿鲁比奥 】 据玉渊谭天,当地时间2月13日,中共中央政治局委员、外交部长王毅在慕尼黑会见美国国务卿鲁比奥。 双重股权结构将赋予特定股东拥有额外投票权的股票,使他们能够主导决策。此举将使马斯克等内部人士即使 持有少 ...
“DeepSeek-V3基于我们的架构打造”,欧版OpenAI CEO逆天发言被喷了
3 6 Ke· 2026-01-26 07:44
Core Viewpoint - The discussion centers around the competitive landscape in the AI field, particularly focusing on the contrasting approaches of Mistral and DeepSeek in developing sparse mixture of experts (MoE) models, with Mistral's CEO acknowledging China's strong position in AI and the significance of open-source models [1][4]. Group 1: Company Perspectives - Mistral's CEO, Arthur Mensch, claims that open-source models are a strategy for progress rather than competition, highlighting their early release of open-source models [1]. - The recent release of DeepSeek-V3 is built on Mistral's proposed architecture, indicating a collaborative yet competitive environment in AI development [1][4]. - There is skepticism among the audience regarding Mistral's claims, with some suggesting that Mistral's recent models may have borrowed heavily from DeepSeek's architecture [4][13]. Group 2: Technical Comparisons - Both DeepSeek and Mistral's Mixtral focus on sparse MoE systems, aiming to reduce computational costs while enhancing model capabilities, but they differ fundamentally in their approaches [9]. - Mixtral emphasizes engineering principles, showcasing the effectiveness of a robust base model combined with mature MoE technology, while DeepSeek focuses on algorithmic innovation to address issues in traditional MoE systems [9][12]. - DeepSeek introduces a fine-grained expert segmentation approach, allowing for more flexible combinations of experts, which contrasts with Mixtral's flat knowledge distribution among experts [11][12]. Group 3: Community Reactions - The community has reacted critically to Mistral's statements, with some users expressing disbelief and pointing out the similarities between Mistral's and DeepSeek's architectures [2][17]. - There is a sentiment that Mistral, once a pioneer in the open-source AI space, is now perceived as having lost its innovative edge, with DeepSeek gaining more influence in the sparse MoE and MLA technologies [14][17]. - The competitive race for foundational models is expected to continue, with DeepSeek reportedly targeting significant releases in the near future [19].
谷歌前CEO施密特:欧洲要么投资开源AI,要么依赖中国模型
Feng Huang Wang· 2026-01-21 07:01
Core Viewpoint - Eric Schmidt emphasizes the necessity for Europe to invest in its own open-source AI labs and address soaring energy prices to avoid dependency on Chinese models [2] Group 1: Investment in AI - Europe must allocate significant funding for developing its own AI models to remain competitive globally [2] - The current trend in the U.S. is towards closed-source AI technologies, which limits flexibility and increases costs [2] Group 2: Comparison with China - China is leading in the development of "open weight" models, which offer higher transparency compared to closed-source models like Google's Gemini and OpenAI's ChatGPT [2] - Without substantial investment, Europe risks relying on Chinese AI models, which could impact its technological sovereignty [2] Group 3: Energy Prices and Infrastructure - High energy prices in Europe pose a challenge for building data centers necessary for training advanced AI technologies [2] - Schmidt expresses concern over the energy demands of AI development in the U.S. and its implications for power supply [2]
外媒热议中国2025年经济亮点
Huan Qiu Shi Bao· 2025-12-31 05:13
Group 1: Economic Resilience and Trade - In 2025, China's goods trade maintained growth for ten consecutive months despite high tariffs imposed by the US, with a predicted export growth rate of 8% for the year [1][2] - China achieved a record annual trade surplus of $1 trillion in November 2025, offsetting declines in exports to specific markets by expanding into Europe, Latin America, and Africa [1] - China's export products have become more innovative, enhancing its role in stabilizing global supply chains amid rising protectionism [2] Group 2: Artificial Intelligence and Technological Advancements - 2025 marked a pivotal year for artificial intelligence, with China's DeepSeek releasing the R1 model, challenging the dominance of US AI companies [3][4] - China is recognized as a leader in the open-source AI sector, with its models gaining traction globally, significantly impacting the competitive landscape [4] - The advancements in AI are part of a broader trend of China's technological capabilities extending into robotics and deep-sea science [3] Group 3: Stock Market Performance - The total market capitalization of A-shares surpassed 100 trillion yuan for the first time in 2025, with significant returns exceeding initial predictions [5][6] - High expectations for the Chinese stock market are driven by a slow bull market and increased foreign investment interest, particularly following the emergence of DeepSeek [5][6] - Analysts predict a 38% increase in the Chinese stock market by the end of 2027, reflecting strong investor confidence [5] Group 4: Soft Power and Global Influence - China ranked second in the global soft power index, surpassing the UK, with cultural products like the toy brand Labubu gaining international popularity [7][8] - The success of Chinese lifestyle brands and cultural products in global markets indicates a shift towards China leading global trends rather than merely following them [7][8] - The rise of Chinese media and entertainment on international platforms showcases the growing influence of Chinese culture [7] Group 5: Consumer Spending and Economic Growth - International organizations have raised China's GDP growth forecast for 2025 to 5%, highlighting its role as a key contributor to global economic growth [9] - Consumer spending's contribution to economic growth significantly increased from 29.7% at the end of 2024 to 56.6% by the third quarter of 2025, driven by both durable goods and service consumption [9] - The outlook for 2026 remains positive, with expectations that consumer spending will continue to support economic growth [9]
中国开源AI逆袭,美国围堵失效,半数美企为何集体倒戈?
Sou Hu Cai Jing· 2025-12-27 06:11
Core Viewpoint - The article discusses the unexpected shift in the U.S. tech landscape, where many American startups are increasingly adopting Chinese open-source AI models despite previous restrictions and concerns about China's AI development [2][10][24]. Group 1: U.S. Companies' Adoption of Chinese AI Models - Over half of U.S. startups are now choosing Chinese open-source AI models as their primary development tools, indicating a significant change in preference [4][10]. - Companies like Perplexity and Airbnb are openly utilizing Chinese models, with Airbnb's CEO stating their AI customer service system heavily relies on Alibaba's Qwen model [6][10]. - The cost-effectiveness of Chinese models is a major factor, with one U.S. entrepreneur noting a switch from a closed-source model that cost $400,000 annually to Qwen, which significantly reduced expenses [10][12]. Group 2: Advantages of Open-Source Models - The annual cost of closed-source models exceeds $1,000 per user, while Chinese open-source models are nearly free, providing a substantial financial incentive for companies [12]. - Open-source models offer greater control and transparency, allowing companies to modify the code as needed without the risk of sudden changes in service terms, as experienced with ChatGPT [12][14]. - The shift from closed to open-source models reflects market dynamics, where companies prioritize economic and security considerations [14][16]. Group 3: Impact of U.S. Restrictions on Chinese AI Development - U.S. restrictions on high-end GPU supplies forced Chinese teams to innovate and optimize algorithms to achieve better performance with limited resources, exemplified by the DeepSeek team [18][20]. - Chinese models are evolving from mere tools to essential infrastructure, similar to the Android system, with millions of developers building applications on these platforms [22][28]. - The competitive edge of Chinese open-source models lies in their low cost, high efficiency, and freedom, challenging the notion that technological progress can be stifled by restrictions [26][29].
英伟达官宣Nemotron 3 新模型,微美全息加码开源AI技术体系革新
Sou Hu Cai Jing· 2025-12-23 06:37
Group 1 - Nvidia has acquired AI software company SchedMD, highlighting its increased investment in open-source technology and the AI ecosystem to address growing competition [1] - SchedMD's core technology, Slurm, is open-source software that assists in scheduling large-scale computing tasks, which significantly occupy data center server capacity [1] - Nvidia's proprietary CUDA software has become an industry standard, driving chip sales and emphasizing the importance of software in maintaining its dominance in the AI field [2] Group 2 - Nvidia has launched the third generation of its "Nemotron" large language model, which boasts faster computation speeds, lower costs, and higher intelligence compared to previous versions, covering various fields such as physical simulation and autonomous driving [2] - The trend in the tech industry is shifting towards open-source AI models, with Nvidia becoming a significant supplier by releasing new models and providing training data and tools for enterprise users [4] - The rapid development of generative AI has led companies to explore key variables for the next decade, including self-developed chips and full-stack software control to accelerate their competitive edge [4] Group 3 - Companies like Weimi Hologram are advancing open-source full-stack AI technology by integrating hardware, software, and ecosystems, aiming to create a comprehensive capability from underlying architecture to industry applications [5] - Weimi Hologram is exploring low-power chips and edge computing solutions to lower the computational barriers for intelligent scenarios, supporting third-party model training through open computing resources [5] - The AI market is rapidly evolving, with a shift towards open-source AI model development, indicating a long-term market potential for various industry applications [6]
逐浪潮 中国大模型跻身第一梯队
Xin Lang Cai Jing· 2025-12-22 23:27
Core Insights - The rise of AI agents is a significant trend in the artificial intelligence sector, marking a shift from tools to partners in various applications, including healthcare and workplace efficiency [1][2][3] Group 1: AI Agents as Partners - AI agents are becoming integral in daily life, assisting with tasks such as analyzing medical reports and generating business insights, thus transforming user interactions with technology [1][2] - The transition from "tool era" to "partner era" in AI is expected to reshape economic structures and human lifestyles, with AI agents facilitating a new interaction paradigm where services proactively reach users [2][3] Group 2: Industry Applications of AI Agents - AI agents are evolving into "smart employees" across industries, contributing to core processes such as predictive maintenance in manufacturing and compliance checks in finance, significantly enhancing operational efficiency [3] - The development of AI agents is supported by advancements in large language models, computational power, and collaborative ecosystems, enabling them to perform complex tasks beyond simple interactions [3] Group 3: Future Trends and Predictions - The concept of "collective intelligence" through multi-agent collaboration is anticipated to become mainstream, allowing dynamic team formations for efficient industrial transformation [4] - The Chinese AI sector is transitioning from a participant to a leader in the global landscape, particularly in open-source AI, showcasing competitive advantages and innovative development paths [5]
小扎千亿新模型被曝“套壳”Qwen,Meta开源已成笑话
3 6 Ke· 2025-12-11 04:04
Core Insights - Meta is experiencing significant internal turmoil as it shifts its AI strategy from open-source models to a more closed approach, particularly with the delayed release of its new model, Avocado, which is reportedly based on the open-source model Qwen [1][2][14]. Group 1: Internal Changes and Strategy Shift - Meta's anticipated release of the Llama 4 model has been disappointing, leading to a strategic pivot towards the Avocado model, which may be closed-source [5][9]. - The company has seen high-profile departures, including its Chief AI Scientist, and has laid off 600 employees as part of a broader restructuring effort [1][4]. - CEO Mark Zuckerberg has implemented aggressive hiring strategies, attracting top talent from the industry to lead the new AI initiatives, breaking from Meta's tradition of promoting from within [13][14]. Group 2: Market Position and Competition - Meta's stock performance has lagged behind the tech sector, raising concerns about its AI strategy and investment returns [18]. - Competitors like OpenAI and Google are advancing rapidly, with OpenAI's ChatGPT and Google's Gemini models gaining significant market traction [20][26]. - The internal confusion regarding the direction of Llama and Avocado reflects a broader uncertainty within Meta about its competitive positioning in the AI landscape [20][21]. Group 3: Future Outlook - The release of the Avocado model has been postponed to Q1 2026 due to ongoing performance testing, indicating a cautious approach to ensure its success [16]. - Meta is also restructuring its AI infrastructure and exploring the integration of various AI models into its products, signaling a shift in focus towards practical applications [25][28]. - The outcome of Meta's AI strategy, particularly with Avocado, is seen as critical for the company's future, potentially determining its trajectory over the next decade [28][29].
这是2025年度AI十大趋势,4个维度10大结论,“开源AI进入中国时间”
Sou Hu Cai Jing· 2025-12-10 15:20
Core Insights - The report highlights a significant shift in AI development from the "tool era" to the "partner era," indicating profound changes in economic structure, social forms, and human lifestyles by 2025 [3][31] Group 1: Key Trends in AI Development - Trend 1: Computing infrastructure is becoming essential, with skyrocketing demand for data centers, marking the computing economy as the primary engine of the intelligent industry [5][6] - Trend 2: AI-driven demand is reshaping chip innovation, with GPUs facing challenges and NPU gaining traction, while ASIC/FPGA are experiencing growth [8][11] - Trend 3: Pre-training will determine the hierarchy of large models, with architectural innovations influencing pre-training levels [13] - Trend 4: Large models are entering the "inference time," with increasing demands for model innovation driven by complex tasks [15] - Trend 5: The period of information AI applications and physical AI research is emerging, with embodied intelligence becoming a focal point [17][20] Group 2: AI Applications and Market Dynamics - Trend 6: AI is reshaping traffic entry points, transitioning from "people finding services" to "services finding people," leading to the evolution of interaction paradigms [20][22] - Trend 7: Multi-modal capabilities are crucial for AI application deployment, enabling systems to process various information types, enhancing productivity [22] - Trend 8: AI hardware is proliferating across devices like PCs, smartphones, and IoT, addressing privacy, latency, and cost efficiency [25] - Trend 9: AI4S is accelerating the realization of AGI, with AI achieving capabilities comparable to doctoral-level problem-solving in various fields [25][27] - Trend 10: Open-source AI is entering a pivotal phase in China, with the country transitioning from a participant to a leader in the global AI landscape [28][30] Conclusion - The report emphasizes that the AI sector is at a historic turning point, with technology evolving from model competition to scenario integration, and highlights China's strategic advancements in open-source ecosystems, autonomous chips, and AGI pathways [31]