作者： Jakub Rusinowski · 最后更新： 2024年8月20日
微软高效的小语言模型（SLM）系列。完美适用于移动设备、边缘计算和低VRAM环境，同时保持基本推理能力。
| Licence | What it permits | Applies to |
|---|---|---|
MIT | Commercial use permitted Commercial use permitted. No usage restrictions beyond attribution. | Phi 3.5 Mini |
| Phi 3.5 Mini | Min 3 GB VRAM · Q4_K_M · 128,000 ctx · ollama run phi3.5 |
The cheapest GPU that runs Phi 3.5 Family locally (min 3 GB VRAM) is the Intel Arc B570 (10 GB).
Install Ollama then run: ollama run phi3.5
Minimum VRAM: 3 GB. For best results use Q4_K_M quantization.
Phi 3.5 Family needs about 3 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Phi 3.5 Mini (3 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Yes — Phi 3.5 Family 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.
Q4_K_M is the best balance of quality and VRAM for Phi 3.5 Family 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.
Install Ollama, then run: ollama run phi3.5. This downloads Phi 3.5 Family and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.