Anthropic今天指控DeepSeek、MiniMax和Moonshot蒸馏Claude的能力这波操作，属实是把贼喊捉贼玩出 了新高度。 我总结成两句话——美国AI商业上卷不动了，就开始掀桌子；技术上又不自信了，就开始搞政治投 机。 1 先把蒸馏这个听起来高大上的技术名词祛魅。说白了，就是AI大模型公司甲付费买了AI大模型公司乙 的API服务，通过格式化提问的方式，拿到了合法的输出结果，然后用这些结果来训练自己的大模型。 先讲业务合理性，在现有的AI竞争环境里，蒸馏本来就是一个司空见惯的手段。现在哪个大模型公司 敢说自己没有蒸馏过别人的大模型？这一直是行业里心照不宣的进化手段和业务规则。 再讲商业，Anthropic自己开门做着API的生意，收着真金白银的调用费，现在我中国AI公司一手交钱， 一手拿货。货（数据）到了我手里，我是拿来做PPT、写代码，还是拿来喂给我的模型做教案，关你屁 事？ 大家都在一个池子里摸鱼，现在中国公司摸了条大鱼，Anthropic就跳脚了。这跟开了家自助餐厅，嫌 客人吃回本了就报警吵着要抓人有什么区别？ 无论在业务还是商业上，中国AI公司根本没有原罪。 2 但更讽刺的是，Anth ...

Core Viewpoint - Mistral, a prominent AI company, is accused of distilling its latest model from DeepSeek, misrepresenting it as a successful reinforcement learning case while distorting benchmark test results [3][21]. Group 1: Company Background - Mistral is recognized as the European version of OpenAI and has gained a strong reputation in the open-source AI community [4][5]. - Founded in 2023 in Paris, Mistral was established by former Google DeepMind and Meta employees [24]. - The company has maintained an open-source approach, releasing models like Mistral Small and Mistral Code, which are competitive in multilingual processing and reasoning capabilities [27]. Group 2: Recent Developments - A recent Twitter leak from a former employee revealed that Mistral's model, Mistral-small-3.2, shows a high similarity to DeepSeek-v3, suggesting possible distillation [12][19]. - The leak indicates that Mistral may have concealed this fact, misleading the public about the effectiveness of its model [21]. - Mistral's valuation reached $10 billion in August 2023, with ongoing fundraising efforts [25]. Group 3: Controversy and Community Reaction - The allegations have sparked controversy, particularly due to Mistral's significant standing in the open-source AI sector [24]. - Many in the community argue that distilled models should be transparently labeled to maintain integrity [22]. - As of now, Mistral has not publicly responded to these allegations, despite recently launching a new model, Mistral Medium V3.1 [29].