Phi-4 Mini

Phi-4 Mini.

Microsoft Released 27 Feb 2025

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Microsoft Feb 2025

Open weights

No Epoch Capabilities Index score yet

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Benchmark	Score	Rank	Source
SIQA	72.5%	#6 of 22	Model card
AlGhafa	69.3%	#12 of 39	Official leaderboard
TruthfulQA	66.4%	#9 of 28	Model card
Open Arabic LLM Leaderboard v2	55.8%	#15 of 39	Official leaderboard
OpenBookQA	79.2%	#8 of 20	Model card
MadinahQA	47.1%	#17 of 39	Official leaderboard
ArabicMMLU	49.5%	#18 of 39	Official leaderboard
ARC-Challenge	83.7%	#22 of 45	Model card
ArbMMLU-HT	42.2%	#20 of 39	Official leaderboard
BoolQ	81.2%	#15 of 28	Model card
ALRAGE	69.1%	#22 of 39	Official leaderboard
BIG-Bench Hard	70.4%	#46 of 71	Model card
Vectara Hallucination Leaderboard (HHEM)	3.40%	#97 of 148	Official leaderboard
Qasper	40.4%	#2 of 3	Model card
MMLU	67.3%	#102 of 127	Model card
FaithJudge	38.4%	#30 of 37	Official leaderboard
MGSM	63.9%	#43 of 52	Model card
PIQA	77.6%	#29 of 33	Model card
WinoGrande	67%	#38 of 42	Model card
MMLU-Pro	52.8%	#137 of 151	Model card
IFEval	70.1%	#116 of 126	Model card
MMMLU	49.3%	#58 of 62	Model card
HellaSwag	69.1%	#44 of 47	Model card
Wer wird Millionär? LLM Benchmark	58.0	#74 of 76	Official leaderboard
Vectara Hallucination Leaderboard (2025 dataset)	23.5%	#103 of 104	Official leaderboard
Arena-Hard-Auto	32.8%	#39 of 39	Model card
SQuALITY	22.8%	#3 of 3	Model card

Benchmark	Score	Rank	Source
GSM8K	88.6%	#45 of 87	Model card
MATH	64%	#49 of 93	Model card
CritPt	0%	#123 of 160	Epoch AI
MATH-500	71.8%	#62 of 67	Model card
GPQA Diamond	30.4%	#277 of 284	Model card
AIME 2024	10%	#85 of 87	Model card

Benchmark	Score	Rank	Source
HumanEval	74.4%	#58 of 95	Model card
MBPP	65.3%	#45 of 66	Model card
BigCodeBench	43%	#9 of 11	Model card
HumanEval+	68.3%	#25 of 27	Model card
StructEval	57%	#26 of 28	Official leaderboard
LiveCodeBench	19.9%	#68 of 73	Model card
MBPP+	63.8%	#14 of 15	Model card
Spider	42.2%	#4 of 4	Model card

Benchmark	Score		Rank	Source
Berkeley Function Calling Leaderboard (BFCL)	70.3%		#22 of 47	Model card

Benchmark	Score	Rank	Source
XSTest	20.8%	#6 of 16	Model card
AraTrust	77.8%	#17 of 39	Official leaderboard
Harmful content defect rate (Azure AI Evaluation SDK, Microsoft)	3.75%	#2 of 3	Model card

The benchmarks the lab chose to headline, with the scores it gave.

Every result in the longer documents the lab published with it. 16 of them aren't in the launch post.