Phi
Microsoft's Phi family of small models trained for strong reasoning at low cost.
Phi is Microsoft's family of small language models, built on the idea that carefully curated training data can give a small model reasoning ability well beyond its parameter count.
The result is models that run cheaply yet handle reasoning and coding tasks surprisingly well, released under a permissive MIT licence.
Where it fits
A fit when you need genuine reasoning from a small, cheap-to-run model with a permissive licence.
Where it does not fit
A small model still has limits: broad world knowledge and the hardest tasks favour a larger one.
Alternatives to Phi
The closest options on the map, and where each fits instead.
-
Gemma Open-weight model family
A fit for self-hosted use cases that need a small, efficient model with the polish of a major lab.
-
SmolLM Open-weight model family
A fit when small size plus full transparency (open data and training recipe) matters for trust or research.
Unlike Phi Maturity: growing
-
Qwen Open-weight model family
A strong fit when you need to size a self-hosted model to specific hardware, or when multilingual coverage matters.
Unlike Phi Hosted API
-
Command R Open-weight model family
A fit when a self-hosted model needs to be good specifically at RAG with citations and tool use.
Pairs with
Commonly used alongside it.
