Liquid LFM3 — Liquid AI 的本地 AI 模型
作者： Jakub Rusinowski · 最后更新：
Liquid AI 第三代边缘模型，于 2026 年 10 月 27 日发布：门控短卷积、滑动窗口注意力和全局注意力的三重混合结构，只有四分之一的层保留全长缓存。文本模型在 4 位精度下约 1 GB。LFM Open License v1.0：年收入低于 1000 万美元可免费商用。
变体
Liquid LFM3 最小的变体在 Q4_K_M 下约需 2 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 |
|---|---|---|---|---|
| LFM3 1.7B → 1.7B | ~1.8 GB | 262,144 | ollama run lfm3:1.7b | |
| LFM3-VL 2B → 2B | ~2 GB | 128,000 | 云端托管标签 |
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 Liquid LFM3
安装 Ollama，然后拉取标签。
ollama run lfm3:1.7b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
Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
适用于: LFM3 1.7B, LFM3-VL 2B推荐 GPU
目录中能在本地运行 Liquid LFM3（至少 2 GB 显存）的最便宜 GPU 是 Intel Arc B570 (10 GB).
Liquid LFM3 — 常见问题
How much VRAM does Liquid LFM3 need?
Liquid LFM3 needs about 2 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: LFM3 1.7B (2 GB, Q4_K_M); LFM3-VL 2B (2 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Liquid LFM3 on an RTX 4090 (24 GB)?
Yes — Liquid LFM3 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 Liquid LFM3?
Q4_K_M is the best balance of quality and VRAM for Liquid LFM3 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 Liquid LFM3 with Ollama?
Install Ollama, then run: ollama run lfm3:1.7b. This downloads Liquid LFM3 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.