Last updated:
Phi
Microsoft's small language models — compact, MIT-licensed models designed to run on consumer hardware.
Model Facts
Latest model: Phi-4-reasoning-vision-15B
- Provider
- Microsoft
- Released
- Mar 1, 2026
- Weights
- Open
Full Model List
| Model | Released | Status | Modalities | Note |
|---|---|---|---|---|
| Phi-4-reasoning-vision-15B | Mar 1, 2026 | current | Text, image | 15B vision-language model that decides when to reason |
| Phi-4 | Dec 12, 2024 | legacy | – | 14B small model |
| Phi-3.5 | Aug 20, 2024 | legacy | – | mini, MoE and vision models |
- Official model page:
- azure.microsoft.com/en-us/products/phi
- Official documentation:
- learn.microsoft.com/azure/ai-foundry
- Other sources:
- Phi (Wikipedia)
What it is
Phi is Microsoft's family of small language models. The Phi-4 generation includes Phi-4 (14B), Phi-4-mini (3.8B), Phi-4-multimodal (5.6B), Phi-4-reasoning and, from March 2026, Phi-4-reasoning-vision-15B, a vision-language model that can decide when to reason. They are MIT-licensed and sized to run on consumer GPUs. No Phi-5 had been released as of this update.
Where this information comes from
Details come from the Phi page on Wikipedia and Microsoft release coverage. The March release date is approximate (month only). Release dates, context windows and benchmark scores are vendor-reported unless stated, and harnesses differ between labs, so treat the numbers as indicative.
Notable traits
- MIT license and small enough for local use.
- Strong reasoning for their size.
- Vision-language and multimodal variants.