Free Phi model via API
Compact Phi models for code and low-latency answers. The table lists hosts currently serving them at no token cost.
2 models1 providerup to 131K contextUpdated:
Read more
Phi wins on latency, not context size. Use another family for long repositories.
Benchmarks: Artificial Analysis (October 24, 2026). A dash means the model is not on their leaderboard; we never estimate scores ourselves.
About Phi
Microsoft · open weights
Microsoft's Phi small language models are open-weight and optimized for reasoning and coding at low latency and cost.
FAQ
Which Phi models are free?
Currently 2 models are free: Phi 3.5 MoE, Phi 3 Vision 128K.
Where can I get a free Phi API key?
Free Phi routes are available at: NVIDIA Build. Get a key in the provider's dashboard; links are on each model card.
Who develops Phi?
Microsoft's Phi small language models are open-weight and optimized for reasoning and coding at low latency and cost.