LFM2.5 — Liquid AI 的本地 AI 模型
作者： Jakub Rusinowski · 最后更新：
Liquid AI's on-device line, built around their hybrid LIV convolution architecture rather than a stack of attention blocks. The 8B-A1B is the flagship: a sparse MoE small enough for phones, laptops and robots, reasoning-only (it emits an explicit chain of thought before answering), and trained on 38T tokens with large-scale RL. Liquid claims tool-calling quality comparable to models up to 4x its active size.
变体
LFM2.5 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 |
|---|---|---|---|---|
| LFM2.5 2.6B → 2.69B | ~2.4 GB | 131,072 | ollama run lfm2-5 | |
| LFM2.5-8B-A1B → 8.3B (1.5B active) | ~5.8 GB | 131,072 | ollama run lfm2.5:8b | |
| LFM2.5-24B-A2B → 24B (2.45B active) | ~15.3 GB | 131,072 | ollama run lfm2.5:24b |
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 LFM2.5
安装 Ollama，然后拉取标签。
ollama run lfm2-5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: LFM2.5 2.6B, LFM2.5-8B-A1B, LFM2.5-24B-A2B推荐 GPU
目录中能在本地运行 LFM2.5（至少 2 GB 显存）的最便宜 GPU 是 Intel Arc B570 (10 GB).
我的 GPU 能运行 LFM2.5 吗？
- LFM2.5 在 AMD Radeon RX 9060 XT 8GB 上
- LFM2.5 在 Intel Arc B570 上
- LFM2.5 在 Intel Arc B580 上
- LFM2.5 在 NVIDIA GeForce RTX 3060 (12GB) 上
- LFM2.5 在 NVIDIA GeForce RTX 3070 上
- LFM2.5 在 NVIDIA GeForce RTX 3070 Ti 上
- LFM2.5 在 NVIDIA GeForce RTX 3080 (10GB) 上
- LFM2.5 在 NVIDIA GeForce RTX 4060 上
- LFM2.5 在 NVIDIA GeForce RTX 4070 上
- LFM2.5 在 NVIDIA GeForce RTX 4070 Super 上
- LFM2.5 在 NVIDIA GeForce RTX 4070 Ti 上
- LFM2.5 在 NVIDIA GeForce RTX 5060 上
- LFM2.5 在 NVIDIA GeForce RTX 5060 Ti 8GB 上
- LFM2.5 在 NVIDIA GeForce RTX 5070 上
LFM2.5 — 常见问题
How much VRAM does LFM2.5 need?
LFM2.5 needs about 2 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: LFM2.5 2.6B (2 GB, Q4_K_M); LFM2.5-8B-A1B (6 GB, Q4_K_M); LFM2.5-24B-A2B (15 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run LFM2.5 on an RTX 4090 (24 GB)?
Yes — LFM2.5 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 LFM2.5?
Q4_K_M is the best balance of quality and VRAM for LFM2.5 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 LFM2.5 with Ollama?
LFM2.5 has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.