Workflow
Composer 2
icon
Search documents
被马斯克认证,Kimi回应
第一财经· 2026-03-21 06:16
2026.03. 21 但很快有开发者发现, Composer 2 是基于 Kimi K2.5 微调而来, 马斯克也在社交媒体发文表 示, " 是的,这就是 Kimi 2.5 。 " Kimi 团队随后以中文热梗 " 听说我,谢谢你,因为有你 " 致谢,被赞 " 温柔中展现技术自信 " 。 | BUND 11075 UT 2 8 1 2 8 1 2 8 1 8 8 1 1 8 5 8 1 1 8 5 8 2 8 1 8 5 8 1 8 5 8 1 8 5 . accounts (anyephere) models, kimi-kilp5-ri-0317-a515-fast | | | --- | --- | | so composer 2 is just Kans 92.5 with RI. at loast rename the model ID | | | 在Curcos: 里便可 Caor AI 的福祉 URL 时刻吸了这个 | | | topper/ts/anyaphere/replace/kima-ki2p5-44-0317-1615-1515-1515 | | | NiS componse 2 赢 ...
Cursor套壳Kimi败露,最强「自研」模型被锤!创始人:忘记署名了
量子位· 2026-03-21 05:11
Core Viewpoint - The article discusses the controversy surrounding the Chinese open-source model Kimi and its alleged misuse by the Silicon Valley company Cursor, highlighting issues of licensing and attribution in the AI model development space [1][2][3]. Group 1: Allegations Against Cursor - The pre-training lead of Kimi, Du Yulun, directly accused Cursor's CEO, Michael Truell, of violating licensing agreements and failing to pay any fees [4][30]. - Co-founder of Kimi, Zhou Xinyu, also criticized Cursor, stating that they did not apply for authorization [6][21]. - The controversy escalated when users discovered that Cursor's new model, Composer 2, was based on Kimi K2.5 without proper attribution [15][26]. Group 2: Cursor's Response and Public Reaction - Cursor admitted that their "self-developed" model was indeed based on Kimi but claimed it was an oversight not to credit Kimi in their release [11][51]. - The public reaction was overwhelmingly negative, with many users expressing disbelief that a company of Cursor's size would neglect to acknowledge the source of their model [20][52]. - The situation intensified as users pointed out that Cursor's performance claims were misleading, as they had not provided transparent details about their model's training process [59][60]. Group 3: Licensing and Market Implications - Kimi's licensing terms require attribution if the model is used in commercial products that exceed certain thresholds, which Cursor clearly surpassed with a valuation of $50 billion and an annual recurring revenue of $2 billion [21][22]. - The article suggests that the incident reflects broader industry practices where smaller companies often rely on established open-source models to reduce costs and risks associated with developing proprietary models from scratch [64][71]. - The controversy serves as a reminder of the importance of transparency and ethical practices in the rapidly evolving AI landscape, especially as companies navigate the complexities of open-source licensing [63][66].
Cursor套壳Kimi被抓包记
机器之心· 2026-03-21 03:27
然而,就在 Composer 2 公开问世后不到 3 小时,就有 X 网友发现 Composer 2 的模型 ID 是 kimi-k2p5-rl-0317-s515-fast,他得出结论说:「 所以 Composer 2 其实 就是经过强化学习的 Kimi K2.5 」。 编辑|Panda 昨天,明星 AI 编程创企 Cursor 发布了自家的新一代模型 Composer 2,其在 CursorBench 基准上的性价比表现明显超过了 Claude Opus 4.6 (high) 以及 GPT-5.4 (high)。 来自 X @fynnso 此事引发了 AI 社区的高度关注,一时之间,梗图与迷因齐飞: 左右滑动查看 就连 X 之王兼世界首富与特斯拉和 SpaceX 大当家伊隆・马斯克也来凑了个热闹。 当然,网友关注的重点并不是 Cursor 竟然是基于 Kimi K2.5 后训练得到了 Composer 2,毕竟基于开源模型进行后训练其实也算是 AI 领域的一种常规操作。大家讨 论的重点是 Cursor 在介绍 Composer 2 时对这个来自中国的开源模型只字未提。 上下滑动查看, 截至 21 日上午截 ...
Cursor自研模型反超Opus 4.6!价格脚踝斩,氛围编程沸腾了
量子位· 2026-03-20 03:52
一水 发自 凹非寺 量子位 | 公众号 QbitAI 众所周知,Cursor作为模型提供商,早期还靠供应Claude模型狠狠吸了一波粉。 结果现在,它自己搞出了一款编程模型,而且转身就把Claude拉下马了—— 其最新编程模型 Composer 2 ,不仅能力超越Claude Opus 4.6,关键是价格降了非常多。 就这么说吧,别人降价是"腰斩",它这直接是 "脚踝斩" 。 那么问题来了,Cursor凭啥能在大家都"涨价"的时候把价格打下去呢? 倒反天罡了朋友! Cursor新模型不仅性能超越Claude,而且价格更是直接"脚踝斩" (都不说腰斩了) 。 鉴于"龙虾"爆火后编程消耗的Token用量一路激增,所以Cursor当下只有一个目标—— 答案,Cursor也随之公布了—— 一种新的强化学习方法 。 比Opus 4.6更强,价格还down down down! 先说目前已经在Cursor上线的 Composer 2 。 从名字英译"编曲家"你就能猜出来了,这款模型主打的是"编程家"(bushi。 性价比、性价比、还是性价比。 何谓性价比?自然是"兼顾智能与成本的最优组合"。 (注:随着"龙虾"爆火,全球 ...