Phi
Microsoft's small open-weight models, strong at reasoning for their size.
Overview
Phi is Microsoft's family of small open-weight language models. They're trained heavily on high-quality and synthetic data, so they do well at reasoning, maths and coding for their size, and they run on a single GPU or even a laptop.
Recent Phi models, including Phi-4, are released under the MIT licence, which allows commercial use.
Models in this family
| Model | Released | Parameters | Context |
|---|---|---|---|
| Phi-4 | 12 Dec 2024 | 14.7B | 16K tokens |
Where to host Phi models
All hosts →| LLM host | Models | Hosting plans |
|---|---|---|
| xCloud | Phi-4 | xCloud LocalAI Hosting, xCloud Ollama Hosting |
News about Phi
All news →xCloud joins LLM Hosting Pro with managed Ollama and LocalAI plans
We've added xCloud as a host, with two managed CPU plans for running open-weight models on Ollama or LocalAI. We've also added Microsoft's Phi family to the site, starting with Phi-4.

