AI Models / Compare

Phi-4

Best-in-class 14B SLM for STEM reasoning and coding; fits consumer hardware.

Creator: Microsoft
Lifecycle: Active
Context: 16.4K
Max output: 16.4K
Released: Dec 12, 2024
Status: unknown

Input: $0.13 / 1M tokens
Output: $0.52 / 1M tokens
Cached read: — / 1M tokens
Cached write: — / 1M tokens
Batch discount: —%
Source: Phi-4 pricing
Verified: Apr 5, 2026 (High)

Capabilities

Modalities: text→text
Capabilities: batchSupportfunctionCallingstructuredOutputs
Strengths: Beats models 3× its size on STEM, Fits in 16 GB RAM
Tradeoffs: Short 16K context, text-only

Official Links

Intro

https://azure.microsoft.com/en-us/blog/introducing-phi-4-microsoft-s-newest-small-language-model-specializing-in-complex-reasoning/

Docs

https://azure.microsoft.com/en-us/products/phi

Pricing

https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/announcing-new-phi-pricing-empowering-your-business-with-small-language-models/4395112

Release note

https://azure.microsoft.com/en-us/updates/?query=phi

Benchmark Coverage

Benchmark	Version	Score	Date	Source	Notes
MMLU	2024	84.8 %	Dec 1, 2024	Microsoft	Vendor-reported
MATH	2024	80.4 %	Dec 1, 2024	Microsoft	Vendor-reported
HumanEval	2024	82.6 %	Dec 1, 2024	Microsoft	Vendor-reported

Release History

Release	Alias	Lifecycle	Release Date	Deprecation	Shutdown	Summary
Phi-4	phi-4	Active	Dec 12, 2024	—	—	Current published model family snapshot.

Host Coverage

Host	Type	Context	Pricing Note	Differences
Azure AI Foundry	first-party	16.4K	$0.13/$0.52 per MTok.	—

Migration Guidance

Best open-weight SLM for constrained compute. Upgrade to Phi-4 MoE or Llama 3.x for larger context.

Change Events

Date	Type	Title	Description	Source
Dec 12, 2024	family_added	Phi-4 published	Initial public model family launch.	Phi-4 release notes

Other models from Microsoft