GPT-4.1 mini

GPT-4.1 mini places in the top three on 3 benchmarks.

OpenAI Released April 14th, 2025 Compare with others

119

Benchmarks with a score

Ranked first, out of five or more models

In the top three

OpenAI Apr 2025

GPT-4.1 mini

Closed weights

134.2 #144 of 256

Epoch Capabilities Index

Self-Reported Index 126.5 #305 of 404

GPQA Diamond 65.8% #126 /229
FrontierMath 4.48% #41 /55
ProfBench 42.8% #54 /72

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.

Knowledge and language

Benchmark	Score	Rank	Source
Vectara Hallucination Leaderboard (HHEM)	2.20%	#27 of 148	Official leaderboard
MMLU	87.5%	#33 of 131	Lab model card
Phare	70.5%	#20 of 62	Official leaderboard
FaithJudge	15.7%	#12 of 37	Official leaderboard
MMBench-CN v1.1	78.6%	#59 of 178	Official leaderboard
LMArena (Elo)	1,383	#152 of 335	Official leaderboard
Arena-Hard v2.0	46.9%	#12 of 23	Official leaderboard
Wer wird Millionär? LLM Benchmark	2489.0	#43 of 77	Official leaderboard
RewardBench 2	65.7%	#24 of 42	Official leaderboard
Internal API instruction following (OpenAI)	45.1%	#5 of 8	Official leaderboard
Multi-IF	67%	#18 of 28	Lab model card
MMMLU	78.5%	#46 of 71	Lab model card
IFEval	84.1%	#90 of 135	Lab model card
HELM Arabic	76.8	#19 of 28	Official leaderboard
EQ-Bench 3	1,097	#53 of 78	Official leaderboard
OpenAI MRCR	47.2%	#7 of 10	Lab model card
SimpleQA	16.8%	#12 of 16	Official leaderboard
Judgemark v2.1	57.3	#33 of 42	Official leaderboard
Creative Writing v3	55.5	#95 of 120	Official leaderboard
GraphWalks	61.7%	#22 of 27	Lab model card
Fiction.LiveBench	44.4%	#47 of 57	Fiction.live leaderboard, via Epoch AI
COLLIE	54.6%	#10 of 12	Lab model card
MARS-Bench	31.2%	#13 of 15	Official leaderboard
AMemGym	36.7%	#10 of 11	Official leaderboard
MultiChallenge	35.8%	#39 of 42	Lab model card
SimpleQA Verified	12.7%	#88 of 94	Epoch AI run
Seal-0	0%	#11 of 11	Official leaderboard

Math and science

Benchmark	Score	Rank	Source
MATH Level 5	87.3%	#20 of 94	Epoch AI run
GPQA Main	61.4%	#23 of 60	Official leaderboard
Misguided Attention	48.7%	#20 of 45	Official leaderboard
SciArena	1,015	#16 of 34	Official leaderboard
GPQA Diamond	65.8%	#126 of 229	Epoch AI run
PolyMath	28.8%	#22 of 38	Official leaderboard
DTBench	68.8%	#116 of 180	Official leaderboard
OTIS Mock AIME 2024-2025	44.7%	#138 of 212	Epoch AI run
MATH-500	88%	#27 of 38	Official leaderboard
FrontierMath	4.48%	#41 of 55	Official leaderboard
AIME 2024	49.6%	#75 of 88	Lab model card
LiveMathematicianBench	20.4%	#16 of 17	Official leaderboard
FrontierMath Tiers 1-3 v2	6.67%	#82 of 86	Official leaderboard
ARC-AGI-2	0%	#95 of 98	Official leaderboard
ARC-AGI	3.50%	#93 of 95	Official leaderboard

Coding

Benchmark	Score	Rank	Source
BigCodeBench-Hard	31.8%	#14 of 96	Official leaderboard
StructEval	76.1%	#11 of 28	Official leaderboard
BigCodeArena	1,000	#9 of 13	Official leaderboard
WeirdML	37.6%	#87 of 112	Official leaderboard
DesignArena Game Dev	1,061	#114 of 143	Official leaderboard
SWE-bench Verified	23.9%	#51 of 60	Official leaderboard
Aider Polyglot	34.7%	#26 of 30	Lab model card
DesignArena Website	990	#133 of 151	Official leaderboard
DesignArena UI Component	952	#127 of 139	Official leaderboard
DesignArena Code	984	#132 of 143	Official leaderboard
DesignArena 3D Design	819	#136 of 136	Official leaderboard

Agents and tools

Benchmark	Score	Rank	Source
TextQuests	11.7	#3 of 28	Official leaderboard
Galileo Agent Leaderboard v2	56%	#3 of 21	Official leaderboard
Berkeley Function Calling Leaderboard (BFCL)	65.2%	#13 of 67	Official leaderboard
Galileo Agent Leaderboard	0.91	#7 of 29	Official leaderboard
BFCL V4	50.5%	#21 of 65	Official leaderboard
SnakeBench	25	#82 of 188	Official leaderboard
ComplexFuncBench	49.3%	#2 of 3	Lab model card
LLM Chess	-167	#91 of 123	Official leaderboard
CompileBench	66.7%	#15 of 20	Official leaderboard
AppWorld Test-Challenge	37.6%	#9 of 12	Official leaderboard
LisanBench	126.3	#110 of 133	Official leaderboard
AppWorld Test-Normal	61.9%	#10 of 12	Official leaderboard
Mystery Game Puzzles	7.00%	#77 of 84	Epoch AI run
tau-bench	45.9%	#22 of 23	Lab model card
MCPMark	3.94%	#27 of 28	Official leaderboard
Chess Puzzles	1.00%	#92 of 93	Epoch AI run
tau2-bench	53%	#10 of 10	Official leaderboard
Cooperate to Compete (C2C)	-1.97	#5 of 5	Official leaderboard

Professional work

Benchmark	Score	Rank	Source
LEXam	54.6%	#9 of 30	Official leaderboard
MortgageTax	65.5%	#38 of 87	Official leaderboard
TaxEval v2	71.9%	#67 of 129	Official leaderboard
CorpFin v2	57.9%	#81 of 119	Official leaderboard
Vals Finance Agent v1	20.8%	#32 of 47	Official leaderboard
LegalBench	78%	#96 of 138	Official leaderboard
ProfBench	42.8%	#54 of 72	Official leaderboard
EnterpriseClawBench	28.6%	#8 of 9	Official leaderboard
CadEval	16%	#12 of 13	Official leaderboard
PRBench	30.4%	#47 of 48	Official leaderboard

Vision, audio and robotics

Benchmark	Score	Rank	Source
SEED-Bench-2-Plus	71.9%	#4 of 136	Official leaderboard
MMBench	86.9%	#10 of 155	Official leaderboard
CharXiv (DQ)	88.4%	#5 of 67	Official leaderboard
MM-Vet	74.3%	#13 of 164	Official leaderboard
A-Bench	77.4%	#14 of 116	Official leaderboard
MMBench-EN v1.1	83.2%	#30 of 178	Official leaderboard
VideoEval-Pro	39.9%	#2 of 10	Official leaderboard
CCBench (MMBench Chinese Culture)	64.3%	#31 of 152	Official leaderboard
VHELM	65.6	#11 of 41	Official leaderboard
MMStar	60.9%	#55 of 173	Official leaderboard
Math-VR	33.3%	#13 of 26	Official leaderboard
MMMU	72.7%	#54 of 103	Lab model card
MathVista	73.1%	#40 of 70	Lab model card
Vision Arena	1,201	#78 of 136	Official leaderboard
VISTA	41.1%	#32 of 54	Official leaderboard
vCX Benchmark	82.9%	#29 of 42	Official leaderboard
MADQA	53%	#14 of 18	Official leaderboard
MathCanvas-Bench	14.6%	#13 of 16	Official leaderboard
CharXiv	56.8%	#67 of 75	Lab model card

Safety and security

Benchmark	Score	Rank	Source
SimpleSafetyTests	100%	#11 of 75	Official leaderboard
BioTIER Permit	99.8%	#13 of 69	Official leaderboard
HELM Safety	94.9	#22 of 75	Official leaderboard
Anthropic Red Team	99.3%	#27 of 75	Official leaderboard
BaxBench	31.8%	#16 of 34	Official leaderboard
AIR-Bench 2024	60.4%	#52 of 82	Official leaderboard
nocot-bench (NCRI)	65.0	#121 of 185	Official leaderboard
Molecular Biology Capabilities Test	42.5%	#67 of 90	Official leaderboard
Human Pathogen Capabilities Test	39%	#71 of 86	Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2)	31.5%	#63 of 71	Official leaderboard
World-Class Bio	24.4%	#67 of 74	Official leaderboard
FORTRESS	59.2%	#67 of 73	Official leaderboard
LiveSecBench	23.0	#32 of 34	Official leaderboard
BioTIER Refuse	10.4%	#66 of 69	Official leaderboard

Other

Benchmark	Score	Rank	Source
OpenVLM Leaderboard	68.9	#40 of 173	Official leaderboard
HELM Capabilities	72.6	#18 of 60	Official leaderboard
LMCA	21.1%	#110 of 147	Official leaderboard
Conceptual Reasoning Index	31.0	#114 of 151	Official leaderboard
ACCoRD	43.6	#125 of 152	Official leaderboard

Its lab's numbers put it at 126.5, 7.7 points below its ECI

The Self-Reported Index fits one score to the 16 capability results OpenAI reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 124.6 to 128.5. How we compute it.

Its lab's numbers next to 6 leaderboards

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark	Lab says	Board says	Gap
IFEval	84.1% IFEval	90.4% Leaderboard, strict, share of instructions followed	−6.3 pts
Aider Polyglot	34.7% Aider's polyglot: whole 34.7 (diff 31.6)	32.4% Leaderboard, aider, diff edits	+2.3 pts
GPQA Diamond	65% GPQA Diamond (model-extracted answers)	65.8% Epoch AI run	−0.8 pts
SWE-bench Verified	23.6% SWE-bench Verified (477 of 500 tasks)	23.9% Leaderboard, mini-SWE-agent	−0.3 pts
Internal API instruction following (OpenAI)	45.1% Internal API instruction following (hard)	45.1% Leaderboard	Same
CharXiv	56.8% CharXiv-R 56.8 (CharXiv-D 88.4)	56.8% Leaderboard	Same

OpenAI's models over time

OpenAI's models in color, against every other model on the Epoch Capabilities Index, by release date.

Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)

Its launch post

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

Introducing GPT-4.1 in the API (OpenAI) Apr 14, 2025

AIME '24 49.6
GPQA Diamond (model-extracted answers) 65
MMLU 87.5
Multilingual MMLU 78.5
SWE-bench Verified (477 of 500 tasks) 23.6
SWE-Lancer (IC-Diamond subset) $31K (13.1%) 13.1
Aider's polyglot: whole 34.7 (diff 31.6) 34.7
Internal API instruction following (hard) 45.1
MultiChallenge (GPT-4o grader) 35.8 (o3-mini grader 42.2) 35.8
COLLIE 54.6
IFEval 84.1
Multi-IF 67
OpenAI-MRCR: 2 needle 128k 47.2 (1M: 33.3) 47.2
Graphwalks bfs <128k 61.7 (parents <128k 60.5) 61.7
MMMU 72.7
MathVista 73.1
CharXiv-R 56.8 (CharXiv-D 88.4) 56.8
ComplexFuncBench 49.3
Mean of Taubench airline 36.0 / retail 55.8 45.9