Hugging Face 推出 SmolLM：超輕量且強大的本地端小模型家族 (135M、360M 與 1.7B)
Original: SmolLM - blazingly fast and remarkably powerful
Hugging Face has officially launched a new family of ultra-lightweight language models called "SmolLM." As generative AI continues to…
Hugging Face 發表全新小語言模型家族 SmolLM，提供 135M、360M 和 1.7B 三種參數規格，專為本地端與行動裝置部署設計。該系列模型在高品質的教育與程式數據集（如 Cosmopedia v2 和 FineWeb-Edu）上進行訓練，效能超越同量級的 MobileLLM 和 Qwen2-0.5B。SmolLM 採 Apache 2.0 開源授權，並同步開源其訓練數據與配方，為邊緣運算與隱私優先的 AI 應用提供極佳選擇。
Hugging Face has officially launched a new family of ultra-lightweight language models called "SmolLM." As generative AI continues to evolve, while large language models (LLMs) perform excellently in the cloud, there is growing demand for "small language models (SLMs)" that can run smoothly on end-user devices like smartphones and laptops — driven by needs for privacy protection, low latency, and offline computation. SmolLM was created precisely to address this pain point, offering three extremely lightweight parameter sizes: 135M (135 million parameters), 360M (360 million parameters), and 1.7B (1.7 billion parameters).
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.