Model

Phi-4-mini-instruct

OPEN

Microsoft · Phi-4 family · released Feb 26, 2025

Compact text-only Phi-4 (3.8B, 128K context) with built-in function calling; tops the 3-4B class, MIT-licensed.

ReasoningCodingVisionFunction callingTool useAgentic

523.2

Elo · rank #298

Parameters

3.8B

Active params

3.8B (dense)

Context

128K tokens

Architecture

Dense decoder-only Transformer (200K vocab, grouped-query attention, shared input/output embedding)

License

MIT

Languages

23+

API price (in/out)

$0.075 / $0.3

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

ARC-ChallengeReasoning83.7%#35

best: Llama 3.1 405B · 96.9%

Arena-HardHuman preference32.8%#51

best: Qwen3-235B-A22B (Non-Thinking) · 96.1%

BIG-Bench HardReasoning70.4%#55

best: ERNIE 4.5 300B-A47B · 94.3%

BFCLAgents70.3%#23

best: SmolLM3 3B (No Thinking) · 92.3%

BigCodeBenchCoding43.0%#6

best: GPT-4o mini · 57.4%

GPQA DiamondReasoning30.4%#262

best: GPT-5.6 · 94.6%

GSM8KMath88.6%#50

best: Llama 3.1 405B · 96.8%

HellaSwagReasoning69.1%#129

best: Claude 3 Opus · 95.4%

HumanEval+Coding68.3%#21

best: Mistral Small 3.2 (24B) · 92.9%

HumanEvalCoding74.4%#67

best: Claude Opus 4.5 · 99.4%

IFEvalReasoning70.1%#114

best: Gemma 4 26B A4B · 98.5%

LiveCodeBenchCoding19.9%#164

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MATH-500Math64.0%#106

best: GPT-5 · 99.4%

MBPP+Coding63.8%#24

best: Llama 3.1 405B · 88.6%

MBPPCoding65.3%#50

best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%

MGSMMath63.9%#39

best: OpenAI o4-mini · 93.7%

MMLU-ProKnowledge52.8%#128

best: Claude Fable 5 · 91.5%

MMLUKnowledge67.3%#166

best: OpenAI o3 · 92.9%

MMMLUKnowledge49.3%#54

best: Gemini 3.1 Pro · 92.6%

PIQAReasoning77.6%#56

best: GPT-4o mini · 93.1%

Social IQaReasoning72.5%#10

best: Apple DCLM-Baseline 7B · 82.9%

TruthfulQAKnowledge66.4%#12

best: Phi-3.5-MoE (16x3.8B, 6.6B active) · 77.5%

WinoGrandeReasoning67.0%#113

best: PaLM 2 · 90.9%

Run it locally

VRAM @ Q4

4 GB

VRAM @ FP16

8 GB

Fits on (Q4)

RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Throughput data unavailable.

Quantizations

GGUF Q4 · MLX 4-bit · ONNX INT4

Phi-4 family

Elo progression across releases

Phi-4 (14B)Dec 20241016.7 Phi-4-mini (3.8B)Feb 2025529.8 Phi-4-mini-instructFeb 2025523.2 Phi-4-multimodal (5.6B)Feb 2025842.4 Phi-4-multimodal-instructFeb 2025842.4 Phi-4-reasoningApr 20251292.1 Phi-4-reasoning (14B)Apr 20251292.1

Resources

HFmicrosoft/Phi-4-mini-instruct GHmicrosoft/PhiCookBook DOCStechcommunity.microsoft.com/blog/educatordeveloperblog/welcome-to-the-new-phi-4-models---microsoft-phi-4-mini--phi-4-multimodal/4386037 APIAvailable via Microsoft API

API price $0.075/$0.3 · each benchmark row carries its own source badge (see methodology)