Phi-4-mini-instruct (3.8B) Benchmark Details

Phi-4-mini-instruct (3.8B) currently shows benchmark results led by GSM8K (16 / 70, score 88.60), HumanEval (47 / 140, score 74.40), MBPP (45 / 96, score 65.30).

Benchmark Results

Phi-4-mini-instruct (3.8B)

Benchmark Results

General Knowledge

3 evaluations

Benchmark / mode

Score

Rank/total

MMLU

Standard Mode

67.30

80 / 124

MMLU Pro

Standard Mode

52.80

134 / 176

HLE

Standard Mode

4.50

500 / 563

Math and Reasoning

5 evaluations

Benchmark / mode

Score

Rank/total

GSM8K

Standard Mode

88.60

16 / 70

MATH-500

Standard Mode

71.80

45 / 45

MATH

Standard Mode

27 / 42

AIME 2024

Standard Mode

60 / 62

AIME2025

Standard Mode

6.70

211 / 215

Coding and Software Engineer

3 evaluations

Benchmark / mode

Score

Rank/total

HumanEval

Standard Mode

74.40

47 / 140

MBPP

Standard Mode

65.30

45 / 96

LiveCodeBench

Standard Mode

12.60

244 / 250

Other

1 evaluations

Benchmark / mode

Score

Rank/total

GPQA Diamond

Standard Mode

439 / 462

Agent Level Benchmark

1 evaluations

Benchmark / mode

Score

Rank/total

τ²-Bench - Telecom

Standard ModeTools

8.20

264 / 264

Instruction Following

1 evaluations

Benchmark / mode

Score

Rank/total

IF Bench

Standard Mode

21.10

278 / 282

AI Agent - Tool Usage

1 evaluations

Benchmark / mode

Score

Rank/total

Terminal-Bench 2.1

Standard ModeTools

0.40

191 / 192

Compare with other models