SmolLM2 — Hugging Face 的本地 AI 模型
作者： Jakub Rusinowski · 最后更新：
最小的生产质量LLM。HuggingFace的SmolLM2模型专为微控制器、手机和通过WebAssembly的浏览器设计。尽管体积极小，通过精心的数据整理展现出令人惊讶的智能。
变体
SmolLM2 最小的变体在 Q4_K_M 下约需 1 GB 显存——量化权重加框架开销，不含 KV 缓存。
| 模型 | Q4 下显存 | 显存 | 上下文 | 运行 |
|---|---|---|---|---|
| SmolLM2 1.7B Instruct → 1.7B | ~1.8 GB | 8,192 | ollama run smollm2:1.7b | |
| SmolLM2 360M Instruct → 360M | ~1 GB | 8,192 | ollama run smollm2:360m |
显存为 Q4_K_M 下的量化权重加开销，与 GPU 与显存检测器使用同一引擎计算。
如何在本地运行 SmolLM2
安装 Ollama，然后拉取标签。
ollama run smollm2:1.7b在上方选择一个尺寸，查看它自己的显存、速度估算和安装命令。
许可证
Commercial use permitted. No usage restrictions beyond attribution.
适用于: SmolLM2 1.7B Instruct, SmolLM2 360M Instruct推荐 GPU
目录中能在本地运行 SmolLM2（至少 1 GB 显存）的最便宜 GPU 是 Intel Arc B570 (10 GB).
我的 GPU 能运行 SmolLM2 吗？
SmolLM2 — 常见问题
How much VRAM does SmolLM2 need?
SmolLM2 needs about 1 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: SmolLM2 1.7B Instruct (2 GB, Q4_K_M); SmolLM2 360M Instruct (1 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run SmolLM2 on an RTX 4090 (24 GB)?
Yes — SmolLM2 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for SmolLM2?
Q4_K_M is the best balance of quality and VRAM for SmolLM2 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run SmolLM2 with Ollama?
Install Ollama, then run: ollama run smollm2:1.7b. This downloads SmolLM2 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.