Phi
A family of small language models (SLMs) developed by Microsoft, specifically engineered to deliver high reasoning capabilities from very small parameter counts.
What it is
The Phi models prove that large scale isn't always necessary. By training on highly curated, "textbook quality" synthetic data, Microsoft created models small enough to run natively on mobile phones that still rival the reasoning skills of models ten times their size.
When you would use it
You use Phi when you are building edge applications or local software where extreme hardware constraints prohibit running standard massive language models.
Common operations
- Embedding AI directly into local mobile or desktop applications without requiring an internet connection.
- Running AI inference on cheap edge devices like Raspberry Pis.
Related terms
Where this is taught
No learning path uses this term yet. Browse the Learning Atlas for guided sequences through related ideas.