Phi-4 Mini

Name: Phi-4 Mini
Author: Microsoft Research

Released

2024-12-13

Last refreshed

2026-06-29

Status

Researched 133d ago

Open sourceCommercial use: permittedLong contextClassification

Microsoft Research releases · 6 in the last 12 months · this family litChangelog →

Specifications

Family: Phi-4
Released: 2024-12-13
Context: 128k
Parameters: 3.8B
Knowledge cutoff: 2024-06
Openness: Open source
License: MITOSI-approvedCommercial use: permitted
Weights: Unknown
Code: Unknown

Created by

Microsoft Research

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

Redmond, Washington, United States

Founded 1991

Website

Pricing

Output / 1M

$0.900

Input / 1M

$0.900

Cheapest of 3 routes · Fireworks AI

Providers(3)

Fireworks AI NVIDIA NIM Novita AI

View 3 provider routes

About

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

Provider price ladder

Compare all 3

Compare API pricing across 2 providers for input and output tokens, batch, and cached reads when available.

Provider	Input / 1M	Output / 1M	Route
Fireworks AI	$0.900	$0.900	Serverless
NVIDIA NIM	-	-	ServerlessPartial

Available via routers & gateways(2)

OpenRouter

Hybrid

Unified hybrid gateway to 400+ models from 60+ providers via a single OpenAI-compatible API, with optional auto-routing that selects the best model per prompt.

PassthroughFireworks AI

NVIDIA LLM Router Blueprint

Router

NVIDIA's open-source AI blueprint for LLM routing that selects the optimal model per prompt via intent classification or neural auto-routing; being deprecated 2026-06-20.

Free OSSNVIDIA NIM

Capabilities

No model capability flags are currently sourced.

Benchmark peer barsfor Classification

MMLU PRORank 91 of 105

Gemini 3 Pro

91.8

Gemini 3.1 Pro Preview

91.0

Qwen3.7-Max

89.6

Claude Opus 4.6

89.1

Phi-4 Minicurrent

52.8

Massive Multitask Language UnderstandingRank 82 of 95

Gemini 3.1 Pro Preview

98.0

92.4

91.1

90.1

67.3

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.

Benchmark	Score	Version	Evaluation	Source
Google-Proof Q&A	25.2	—Observed 2026-04-28	—	Source
Massive Multitask Language Understanding	67.3	—Observed 2026-04-28	—	Source
MMLU PRO	52.8	—Observed 2026-04-28	—	Source