原文标题： Alibaba-backed Moonshot releases its second AI update in four months as China's AI race heats up 原文连结： https://cnb.cx/47Vb7Cy 发布时间：2025-11-06 记者署名：Evelyn Cheng 原文内容： Chinese startup Moonshot on Thursday released its latest generative artificial i ntelligence model which claims to beat OpenAI’s ChatGPT in “agentic” capabili ties — or understanding what a user wants without explicit step-by-step instruc tions. The model, called “Kimi K2 Thinking,” builds on the K2 model released in July by Beijing-based Moonshot, which is backed by Alibaba . The update comes as Nvidia CEO Jensen Huang this week again urged the U.S. to press ahead in a race agains t Chinese-developed AI. Some major U.S. companies such as Airbnb have begun to p ublicly tout how some Chinese AI models are as viable — and often cheaper — al ternatives to OpenAI’s. Despite U.S. restrictions on Chinese businesses’ access to high-end chips, comp anies such as DeepSeek have released AI models that are open sourced and with us er fees a fraction of ChatGPT’s. DeepSeek also claimed it spent $5.6 million for its V3 model — in contrast to t he billions spent by OpenAI. The Kimi K2 Thinking model cost $4.6 million to train, according to a source fam iliar with the matter. It can automatically select 200 to 300 tools to complete tasks on its own, reduc ing the need for human intervention according to Moonshot. CNBC was unable to in dependently verify the DeepSeek or Kimi figures. DeepSeek last month released a new AI model that claims to improve performance b y using visual clues to expand the context of information it is processing at on ce. 中国初创公司月之暗面（Moonshot）周四发布了其最新的生成式人工智慧模型，该模型声称 在「代理能力」方面超越了 OpenAI 的 ChatGPT——即无需明确的逐步指令就能理解用户需 求。 这款名为「Kimi K2 Thinking」的模型建立在总部位於北京的月之暗面公司於 7 月发布的 K2 模型基础之上，该公司获得阿里巴巴的投资支持。 此次更新发布之际，辉达（Nvidia）执行长黄仁勳本周再次敦促美国在与中国开发的 AI 竞 赛中加速前进。包括 Airbnb 在内的一些美国主要企业已开始公开宣传某些中国 AI 模型是 OpenAI 的可行替代方案，且价格通常更为低廉。 尽管美国限制中国企业取得高端晶片，但像 DeepSeek 这样的公司仍发布了开源 AI 模型， 其用户费用仅为 ChatGPT 的一小部分。 DeepSeek 还声称其 V3 模型的开发成本为 560 万美元——这与 OpenAI 投入的数十亿美元 形成鲜明对比。 据知情人士透露，Kimi K2 Thinking 模型的训练成本为 460 万美元。 根据月之暗面的说法，该模型可自动选择 200 至 300 种工具独立完成任务，从而减少人工 干预的需求。CNBC 无法独立核实 DeepSeek 或 Kimi 的相关数据。 DeepSeek 上个月发布了一款新 AI 模型，声称能透过使用视觉线索来扩展其同时处理资讯 的上下文，从而提升性能。 心得： 阿里巴巴投资的moonshot公司 旗下的kimi k2 thinking 在11/6一发布就开源 在某些指标击败了gpt5这样的顶级模型 而且训练成本仅仅460万美金 https://i.imgur.com/kXOblli.jpeg --

※ 发信站: 批踢踢实业坊(ptt.cc), 来自: 42.73.164.44 (台湾) ※ 文章网址: https://webptt.com/cn.aspx?n=bbs/Stock/M.1762675755.A.178.html