Workflow
DeepSeek
icon
Search documents
OpenAI’s ‘code red’ memo lays bare pressure from Google, DeepSeek and its $1.4 trillion AI bet
CNBC Television· 2025-12-02 18:31
Uh McKenzie Seagalos joins us now. What does this what does this mean. I mean, is this now uh put put Google in a in a position now where they have um a a uh an opportunity now to to to beat uh Open AI in any stretch.>> It certainly seems to signal that. So this code red warning comes from a leaked memo cited by the journal and the information and in it Sam Alman tells staff to pause work on ads health and shopping agents and then shift focus back to their core chat GBT experience faster responses better pe ...
OpenAI's ‘code red' memo lays bare pressure from Google, DeepSeek and its $1.4 trillion AI bet
Youtube· 2025-12-02 18:31
Uh McKenzie Seagalos joins us now. What does this what does this mean. I mean, is this now uh put put Google in a in a position now where they have um a a uh an opportunity now to to to beat uh Open AI in any stretch.>> It certainly seems to signal that. So this code red warning comes from a leaked memo cited by the journal and the information and in it Sam Alman tells staff to pause work on ads health and shopping agents and then shift focus back to their core chat GBT experience faster responses better pe ...
好家伙!DeepSeek 一口气连发 2 个新模型
程序员的那些事· 2025-12-02 13:49
转自:量子位 | 公众号 QbitAI 突袭! ChatGPT发布三周年,DeepSeek嚯一下发出两个模型: 前者聚焦平衡实用 ,适用于日常问答、通用Agent任务、真实应用场景下的工具调用。 推理达GPT-5水平,略低于Gemini-3.0-Pro。 后者主打极致推理, 推理基准性能媲美Gemini-3.0-Pro。 还一把斩获IMO 2025、CMO 2025、ICPC World Finals 2025、IOI 2025金牌。 划重点,ICPC达到人类选手第二、IOI人类选手第十名水平。 具体来说,DeepSeek-V3.2侧重于平衡推理能力与输出长度,降低计算开销。 DeepSeek官微推文中写道,"DeepSeek-V3.2模型在Agent评测中达到了当前开源模型的最高水平"。 该模型其他情况如下: 下图展示的是DeepSeek-V3.2与其他模型在各类Agent工具调用评测集上的得分 DeepSeek-V3.2 DeepSeek-V3.2-Speciale 推理能力比肩GPT-5; 相比Kimi-K2-Thinking大幅缩短输出长度,减少用户等待时间; DeepSeek旗下首个"思考融入工具调 ...
Sam Altman Declares Code Red
Seeking Alpha· 2025-12-02 11:57
Listen on the go! A daily podcast of Wall Street Breakfast will be available by 8:00 a.m. on Seeking Alpha, iTunes, Spotify.Getty Images Good morning! Here is the latest in trending:Sweetened offer: Warner Bros. Discovery (WBD) received a mostly cash offer from Netflix (NFLX), which is arranging a bridge loan worth tens of billions of dollars for its bid.Tariff refund: Costco (COST) sued the U.S. government to ensure it gets a full refund of tariffs if the Supreme Court rules against President Trump's levie ...
从开源最强到挑战全球最强:DeepSeek新模型给出了解法
Guan Cha Zhe Wang· 2025-12-02 11:38
Speciale版本是DeepSeek-V3.2的长思考增强版,同时结合了DeepSeek-Math-V2的定理证明能力。目标是将开源模型的推理能力推向极致,探索模型能力的边 界。目前仅以临时API服务形式开放,以供社区评测与研究。 回顾过去一年,开源大模型生态在年初DeepSeek惊艳亮相之后集体爆发,阿里云的Qwen系列不断刷新榜单,月之暗面的Kimi,智谱的GLM和MiniMax的M 系列模型均在发布后收获了国内外的好评并取得了超越当时顶级闭源模型的开源成果。这一波群雄并起的浪潮,将"开源追平乃至超越闭源"从一句口号变成 了让闭源厂商感到压力的现实。 然而,随着Google Gemini3.0的强势发布,凭借庞大的的算力和数据,Gemini 3.0 Pro重新定义了什么是"全球最强"。其强劲的性能甚至让同为竞争对手的马 斯克(xAI)和奥特曼(OpenAI)纷纷点赞,开源和闭源似乎不复存在的差距瞬间又变成了一道新的天花板。 与此同时,OpenAI前首席科学家Ilya Sutskever近期关于"Scaling Law撞墙"的论断,更是给后来者泼了一盆冷水:如果连单纯堆算力都开始失效,那么资源本 就处于劣 ...
中科曙光:曙光AI超集群系统等产品深度适配DeepSeek-V3.2
人民财讯12月2日电,据中科曙光(603019)消息,12月1日,DeepSeek正式发布DeepSeek-V3.2和 DeepSeek-V3.2-Speciale,大幅强化Agent能力,融入思考推理。基于中国首个AI计算开放架构,硬件 层、软件层、模型层实现"跨层协同",曙光AI超集群系统、scaleX640超节点等产品0day完成对 DeepSeek新版本的深度适配与调优,支持各行各业客户进行全量落地部署。 ...
DeepSeek重磅上新,对标美国行业巨头,“所有群聊都炸锅了!”
Xin Lang Cai Jing· 2025-12-02 10:24
[文/观察者网 阮佳琪] 深度求索(DeepSeek)新品重磅发布,再度引爆AI圈。12月1日,中国人工智能(AI)初创企业 DeepSeek推出两款正式版模型:DeepSeek-V3.2和DeepSeek-V3.2-Speciale。 据介绍,DeepSeek-V3.2定位"平衡实用",其在主流推理基准测试中达到美国OpenAI的GPT-5水平;而 在推理能力上大幅增强的DeepSeek-V3.2-Speciale,则在推理基准测试中取得了媲美谷歌深度思维11月 下旬刚推出的新一代AI模型"双子座3专业版"(Gemini 3.0 Pro)。 DeepSeek还透露,其V3.2-Speciale版本在国际数学奥林匹克竞赛(IMO 2025)、国际信息学奥林匹克竞 赛(IOI 2025)等均斩获金牌级表现。这一成绩直接对标行业巨头,此前仅有OpenAI和谷歌深度思维未 对外公开的内部测试模型达成过这一成就。 香港《南华早报》2日报道指出,这一来自开源实验室的技术突破再次引发AI研究领域的广泛热议,尤 其DeepSeek此次上新恰逢"AI界的奥斯卡"——2025年神经信息处理系统大会(NeurIPS)召开前夕。 ...
ChatGPT三周年遭DeepSeek暴击,23页技术报告藏着开源登顶的全部秘密
36氪· 2025-12-02 09:19
DeepSeek V3.2上新黑科技。 来源| APPSO(ID:appsolution) 封面来源 | unsplash ChatGPT诞生三周年之际,DeepSeek送上「庆生礼物」。 12月1日, DeepSeek一口气发布两款模型:DeepSeek-V3.2和DeepSeek-V3.2-Speciale。这两个模型不仅在推理能力上直逼GPT-5和Gemini-3.0-Pro ,更重 要的是,它们解决了一个困扰开源模型很久的问题: 过去几个月,AI圈出现了一个明显的趋势:闭源模型越跑越快,开源模型却有点跟不上节奏了。DeepSeek团队分析后发现,开源模型在处理复杂任务时有 三个核心瓶颈:架构问题、资源分配以及智能体能力。 针对这三个问题,DeepSeek这次拿出了三个大招。 如果你用过一些AI模型处理超长文档,可能会发现速度越来越慢,甚至直接卡死。这就是传统注意力机制的锅。 怎么让AI既会深度思考,又会熟练使用工具? 新模型省流版如下: DeepSeek-V3.2(标准版) :主打性价比与日常使用,推理能力达到GPT-5水平,比Kimi-K2-Thinking输出更短、更快且更省成本,并首次实现「边思 ...
再谈注意力:阿里、Kimi 都在用的 DeltaNet 和线性注意力新改进丨晚点播客
晚点LatePost· 2025-12-02 09:13
以下文章来源于晚点科技 ,作者晚点团队 晚点科技 . 见证奇点来临 不仅是提升效率,线性注意力在数据受限情况下也可能提升效果。 访谈 丨 程曼祺 整理 丨 姚一楠 注意力机制(Attention)是 Transformer 架构大型语言模型(LLM)的核心机制,它决定了模型如何 处理、理解海量的文本信息。然而,传统全注意力机制的计算开销会随文本长度呈平方级暴增,这正 是限制模型处理长文档、长上下文的关键瓶颈。 今年初,《晚点聊》的 103 期和 104 期节目分别讨论了注意力机制改进的两个主要方向:"稀疏注意 力" 和 "线性注意力"。(文字版见《 大模型 "注意力简史":与两位 AI 研究者从 DeepSeek、Kimi 最 新改进聊起 》和《 3700 次预训练寻找 "线性注意力" 非共识,MiniMax-01 开发者讲述 4 年探索 》) 这期节目,我们继续关注线性注意力的新进展。在 9 月和 10 月底,阿里巴巴和月之暗面先后开源 Qwen3-Next 和 Kimi Linear 模型,其中的注意力机制都使用了线性注意力 DeltaNet 和 full attention (传统的全注意力)混合的方式 ...
对标美国行业巨头,“所有群聊都炸锅了”
Guan Cha Zhe Wang· 2025-12-02 08:46
Core Insights - DeepSeek, a Chinese AI startup, has launched two new models, DeepSeek-V3.2 and DeepSeek-V3.2-Speciale, which have achieved performance levels comparable to leading models from OpenAI and Google DeepMind [1][8] - The release of these models coincides with the upcoming NeurIPS conference, generating significant interest in the AI research community [2][8] Model Performance - DeepSeek-V3.2 is designed for practical use, achieving performance on par with OpenAI's GPT-5 in mainstream reasoning benchmarks, while DeepSeek-V3.2-Speciale excels in reasoning capabilities, matching Google DeepMind's Gemini 3.0 Pro [1][4] - The V3.2 model has shown a significant reduction in output length compared to Kimi-K2-Thinking, leading to lower computational costs and reduced user wait times [4] - DeepSeek-V3.2-Speciale has demonstrated exceptional performance in international competitions, including winning gold medals in IMO 2025 and IOI 2025, marking a significant achievement for open-source AI models [5][8] Competitive Landscape - The advancements made by DeepSeek indicate that Chinese open-source AI systems are becoming competitive with top proprietary models from Silicon Valley [8][10] - The trend towards open-source models in China contrasts with the closed strategies of major US tech companies, which tend to keep their advanced AI technologies proprietary [10][11] - Recent data shows that the download share of open-source AI models developed by Chinese teams has surpassed that of US teams for the first time, indicating a shift in the global AI landscape [9][10] Community and Industry Impact - The announcement of DeepSeek's new models has sparked excitement within the AI research community, with discussions and engagement across various platforms [2][8] - The models are now available on DeepSeek's official website, app, and API, with the Speciale version currently offered as a temporary API for community evaluation [5][7]