Phi-4

Microsoft · Phi · open · released Dec 12, 2024 · microsoft/phi-4

Phi family:Phi-3.5 Mini·Phi-4

Microsoft's 14B small model — exceptional reasoning quality per parameter, competitive with much larger models.

Pricing

InputOpen weights / $0.07 / 1M via Azure
OutputOpen weights / $0.14 / 1M via Azure
Cost bandFree/open
Free tierOpen weights (MIT); hosted via Azure AI Foundry.

Limits

Context window16k checked Jul 19, 2026
Max outputRuntime-dependent
Knowledge cutoffJun 2024
DeploymentAzure AI Foundry, Hugging Face, Ollama, self-hosted
OpenAI-compatible APIYes
ModalitiesText

Where it fits

Best forOn-premise or resource-constrained setups where reasoning quality must be maximized in a 14B footprint.
Business fitExcellent for edge AI and enterprise use cases with strict hardware budgets.

Capabilities

Open weights (MIT)ReasoningMathCode

Benchmark scores

Composite74 / 100
Coding76 / 100 (HumanEval)
Reasoning78 / 100 (MMLU-Pro)
Long context55 / 100
Instruction following78 / 100 (IFEval)

Editorial aggregate of published benchmark results as of Jan 1, 2025. These are not independent measurements run by this site.

Change history

No recorded changes for this model yet. Pricing and context are checked daily; changes appear here with dates and sources.

Data

Last checkedJul 19, 2026
Cross-checked againstopenrouter.ai/microsoft/phi-4

Verify with provider documentation before committing spend. Report an error and it gets fixed with a logged correction.

Open in the comparison table JSON API