以 DeepSeek 自己做的蒸馏尝试为例：基于隔壁千问蒸馏自家的 R1 模型后得到的 DeepSeek-R1-Distill-Qwen 1.5B 这个小模型，仅靠 7000 条样本和极低的计算成本，就在 AIME24 数学竞赛基准上超越了 OpenAI 的 o1-preview。
Lex: FT's flagship investment column
，推荐阅读im钱包官方下载获取更多信息
This article originally appeared on Engadget at https://www.engadget.com/ai/ai-robotics-company-started-by-alphabet-is-joining-google-proper-144421411.html?src=rss。Line官方版本下载对此有专业解读
SpaceX has not responded to emailed requests from BBC News for comment. The researchers also sent their findings to the company but did not receive a response.