LLM ReferenceLLM Reference

Phi-4 Mini

phi-4-mini

Researched 137d ago

Last refreshed 2026-05-11. Next refresh: weekly.

Open SourceClassification

Phi-4 Mini is worth evaluating for classification when its provider route and context window match the workload.

Decision context: Classification task fit, 3 tracked provider routes, and research from 2026-01-01.

Use it for

  • Teams evaluating classification
  • Buyers comparing 3 tracked provider routes

Do not use it for

  • Vision or document-understanding workloads
  • Strict JSON or tool-calling flows

Cheapest output

$0.150

Novita AI per 1M tokens

Provider routes

3

Tracked API hosts

Quality / dollar

Grade B

Ranked by benchmark score divided by cheapest output price

Freshness

2026-01-01

Researched 137d ago

stale

Top use-case fit

Classification

Q/$ B

2 relevant benchmarks in the decision map.

Provider price ladder

ProviderInput / 1MOutput / 1MRoute
Novita AI$0.050$0.150
Serverless
Fireworks AI$0.900$0.900
Serverless
NVIDIA NIM--
ServerlessPartial

Benchmark peer barsfor Classification

Migration checks

No linked migration route is available for this model yet.

About

Phi-4 family model from Microsoft Research. Mini variant with efficient performance.

Phi-4 Mini input tokens at $0.05/1M, output at $0.15/1M.

Capabilities

No model capability flags are currently sourced.

Benchmark Scores(3)

Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.
BenchmarkScoreVersionSource
Google-Proof Q&A25.2https://huggingface.co/microsoft/Phi-4-mini-instruct
Massive Multitask Language Understanding67.3https://huggingface.co/microsoft/Phi-4-mini-instruct
MMLU PRO52.8https://huggingface.co/microsoft/Phi-4-mini-instruct

Rankings

Specifications

FamilyPhi-4
Released2024-12-13
Parameters3.8B
LicenseMicrosoft Research

Created by

Advancing the state-of-the-art in AI and computing.

Redmond, Washington, United States
Founded 1991
Website