「DeepSeek 時刻」一週年：開源 AI 的典範轉移與變革回顧
Original: One Year Since the “DeepSeek Moment”
The DeepSeek-V3 and R1 models released in January 2025 have been hailed as the "DeepSeek Moment" in the AI world. This upheaval not only…
2025 年初的「DeepSeek 時刻」以超低訓練成本與強大推理能力震驚全球。一年過去，Hugging Face 回顧了這場變革對開源社群的深遠影響，包括 GRPO 強化學習演算法的普及、MoE 架構的廣泛應用，以及全球開發者如何擺脫對閉源巨頭的依賴，走向更自主、高效的本地部署與微調時代。這標誌著 AI 發展從「算力軍備競賽」走向「演算法與效率至上」的新紀元。
The DeepSeek-V3 and R1 models released in January 2025 have been hailed as the "DeepSeek Moment" in the AI world. This upheaval not only shattered the myth that "only trillion-dollar giants can build frontier models," but also reshaped the entire AI industry through an open-source approach and an remarkably low training cost of approximately $5.6 million.
Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.
See Pro plans →Want the original English / full article?
Read on Hugging Face Blog →Summaries are AI-generated; the original article is authoritative.