GPT-5 nano

GPT-5 nano ranks first on HELM Safety.

OpenAI Released August 7th, 2025 Compare with others

118

Benchmarks with a score

Ranked first, out of five or more models

In the top three

OpenAI Aug 2025

GPT-5 nano

Closed weights

139.1 #128 of 256

Epoch Capabilities Index

Self-Reported Index 133.4 #245 of 404

MMMU-Pro 62.6% #64 /90
Terminal-Bench 2.0 21.8% #39 /43
ZeroBench 3.00% #60 /76

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
IFEval	93.2%	#4 of 60	Official leaderboard
COLLIE	96.9%	#3 of 12	Lab model card
SimpleQA Hallucination Rate	31%	#3 of 12	Lab model card
MMBench-CN v1.1	79.6%	#48 of 178	Official leaderboard
Slop Score	34.1	#6 of 17	Official leaderboard
Phare	69.5%	#22 of 62	Official leaderboard
SuperGPQA	48.9%	#24 of 59	Official leaderboard
Wer wird Millionär? LLM Benchmark	13156.0	#32 of 77	Official leaderboard
SnorkelSpatial	26.7%	#12 of 27	Official leaderboard
Internal API instruction following (OpenAI)	56.1%	#2 of 4	Lab model card
LongFact-Concepts	0.90%	#2 of 4	Lab model card
LMArena (Elo)	1,338	#190 of 335	Official leaderboard
LongFact	1.00%	#3 of 5	Lab model card
Vectara Hallucination Leaderboard (2025 dataset)	10.5%	#63 of 104	Official leaderboard
Vectara Hallucination Leaderboard (HHEM)	4.70%	#90 of 148	Official leaderboard
MultiChallenge	54.9%	#26 of 42	Lab model card
GraphWalks	64%	#19 of 27	Lab model card
LongFact-Objects	3.10%	#3 of 4	Lab model card
SimpleQA	11%	#48 of 60	Lab model card
OpenAI MRCR	43.2%	#8 of 10	Lab model card
Fiction.LiveBench	44.4%	#49 of 57	Fiction.live leaderboard, via Epoch AI
Humanity's Last Exam	8.70%	#144 of 154	Lab model card
Creative Writing v3	45.9	#114 of 120	Official leaderboard
SimpleQA Verified	11.7%	#90 of 94	Epoch AI run
Longform Creative Writing (EQ-Bench)	28.0	#115 of 119	Official leaderboard
BrowseComp Long Context 128k	80.4%	#3 of 3	Lab model card
FActScore	7.30%	#3 of 3	Lab model card

Math and science

Benchmark	Score	Rank	Source
SnorkelSequences	72%	#3 of 25	Official leaderboard
MATH Level 5	95.2%	#12 of 94	Epoch AI run
GPQA Main	67.9%	#13 of 60	Official leaderboard
ChemBench	63.4%	#8 of 24	Official leaderboard
MATH-500	93.8%	#16 of 38	Official leaderboard
OTIS Mock AIME 2024-2025	81.1%	#96 of 212	Epoch AI run
CMT-Benchmark	14%	#5 of 11	Official leaderboard
SnorkelGraph	62.5%	#11 of 24	Official leaderboard
SMT 2025	85.4%	#19 of 40	Official leaderboard
AIME 2025	85.2%	#72 of 143	Lab model card
GPQA Diamond	69.4%	#120 of 229	Epoch AI run
FrontierMath	8.28%	#34 of 55	Official leaderboard
CMIMC 2025	74.4%	#22 of 33	Official leaderboard
ProofBench	12%	#41 of 59	Official leaderboard
DTBench	62.7%	#133 of 180	Official leaderboard
HMMT 2025	75.6%	#41 of 54	Lab model card
FrontierMath Tier 4	2.08%	#42 of 55	Epoch AI run
ARC-AGI-2	2.61%	#76 of 98	Official leaderboard
ARC-AGI	20.7%	#81 of 95	Official leaderboard
SciArena	903	#29 of 34	Official leaderboard
HMMT Nov 2025	81.7%	#19 of 22	Official leaderboard
FrontierMath Tiers 1-3 v2	20%	#75 of 86	Official leaderboard
BRUMO 2025	80.8%	#37 of 42	Official leaderboard
FrontierMath Tier 4 v2	2.44%	#62 of 68	Official leaderboard
ForecastBench	0.17	#75 of 82	Official leaderboard
MaCBench	68.8%	#2 of 2	Official leaderboard

Coding

Benchmark	Score	Rank	Source
Aider Polyglot	48.4%	#22 of 30	Lab model card
WeirdML	38.1%	#86 of 112	Official leaderboard
SWE-bench Verified	34.8%	#48 of 60	Official leaderboard
ALE-Bench	1,016	#105 of 126	Official leaderboard

Agents and tools

Benchmark	Score	Rank	Source
SnakeBench	28	#54 of 188	Official leaderboard
LLM Chess	683	#37 of 123	Official leaderboard
BFCL V4	51.5%	#20 of 65	Official leaderboard
LisanBench	1217.7	#52 of 133	Official leaderboard
AdaPlanBench	42.4%	#4 of 10	Official leaderboard
LLM-WikiRace	4.00%	#11 of 21	Official leaderboard
Chess Puzzles	22%	#62 of 93	Epoch AI run
SnorkelWordle	79%	#13 of 19	Official leaderboard
Terminal-Bench	21.8%	#14 of 18	tbench.ai, via Epoch AI
Mystery Game Puzzles	9.00%	#72 of 84	Epoch AI run
MCPMark	6.30%	#25 of 28	Official leaderboard
Terminal-Bench 2.0	21.8%	#39 of 43	Official leaderboard
PinchBench	41.5%	#44 of 48	Official leaderboard
tau2-bench	46.3%	#74 of 79	Lab model card

Professional work

Benchmark	Score	Rank	Source
SnorkelUnderwrite	47%	#17 of 27	Official leaderboard
TaxEval v2	67.4%	#99 of 129	Official leaderboard
MedScribe	72.9%	#77 of 97	Official leaderboard
CaseLaw v2	52.6%	#43 of 53	Official leaderboard
FinQA (Snorkel OpenEnv environment)	50%	#18 of 22	Official leaderboard
SnorkelFinance	50%	#18 of 22	Official leaderboard
MortgageTax	53.6%	#73 of 87	Official leaderboard
LEXam	27.3%	#26 of 30	Official leaderboard
MedCode	30.4%	#89 of 96	Official leaderboard
ProfBench	36.9%	#68 of 72	Official leaderboard
Finance Reasoning (Snorkel)	26.6%	#20 of 21	Official leaderboard
LegalBench	50.1%	#136 of 138	Official leaderboard

Vision, audio and robotics

Benchmark	Score	Rank	Source
MMStar	68.3%	#22 of 173	Official leaderboard
MM-Vet	67.4%	#28 of 164	Official leaderboard
MMBench-EN v1.1	78.9%	#64 of 178	Official leaderboard
MMMU	75.6%	#39 of 103	Lab model card
MMMU-Pro	62.6%	#64 of 90	Lab model card
MADQA	58.2%	#13 of 18	Official leaderboard
Vision Arena	1,147	#100 of 136	Official leaderboard
VPCT	37.2%	#20 of 27	VPCT leaderboard, via Epoch AI
ERQA	50.1%	#27 of 36	Lab model card
ZeroBench	3.00%	#60 of 76	Official leaderboard
CharXiv	62.7%	#62 of 75	Lab model card
Video-MMMU	66.8%	#31 of 36	Lab model card
RoboRewardBench	1.30	#20 of 22	Official leaderboard
Video-MME	65.7%	#47 of 51	Lab model card
ClockBench	3.90%	#36 of 37	Official leaderboard
MineBench	563	#75 of 75	Official leaderboard
vCX Benchmark	56.4%	#42 of 42	Official leaderboard

Safety and security

Benchmark	Score	Rank	Source
HELM Safety	98.6	#1 of 75	Official leaderboard
AIR-Bench 2024	87.8%	#6 of 82	Official leaderboard
Anthropic Red Team	99.5%	#17 of 75	Official leaderboard
SimpleSafetyTests	100%	#21 of 75	Official leaderboard
KORA	48.4%	#20 of 42	Official leaderboard
Human Pathogen Capabilities Test	45.4%	#59 of 86	Official leaderboard
World-Class Bio	33.6%	#56 of 74	Official leaderboard
Molecular Biology Capabilities Test	41%	#73 of 90	Official leaderboard
nocot-bench (NCRI)	31.0	#182 of 185	Official leaderboard

Other

Benchmark	Score	Rank	Source
OpenVLM Leaderboard	72.2	#23 of 173	Official leaderboard
HELM Capabilities	74.8	#13 of 60	Official leaderboard
ACCoRD	58.8	#94 of 152	Official leaderboard
CAIS Risk Index	55.6	#43 of 67	Official leaderboard
Conceptual Reasoning Index	24.1	#130 of 151	Official leaderboard
LMCA	7.93%	#137 of 147	Official leaderboard
SAGE (Vals)	30.4%	#76 of 81	Official leaderboard
CAIS Vision Capabilities Index	34.3	#51 of 53	Official leaderboard
CAIS Text Capabilities Index	6.60	#68 of 68	Official leaderboard

Its lab's numbers put it at 133.4, 5.7 points below its ECI

The Self-Reported Index fits one score to the 18 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 130.8 to 136.0. How we compute it.

Its lab's numbers next to 5 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	What differs
SWE-bench Verified	54.7% SWE-bench Verified (477 of 500 tasks)	34.8% Leaderboard, medium, mini-SWE-agent	+20 pts	The lab evaluates 477 tasks, while the board reports a mini-SWE-agent run at medium reasoning effort.
GPQA Diamond	71.2% GPQA diamond (no tools)	69.4% Epoch AI run, high	+1.8 pts
HMMT 2025	75.6% HMMT 2025 (no tools)	74.2% Leaderboard, high	+1.4 pts
FrontierMath	9.60% FrontierMath (with python tool only)	8.28% Leaderboard, high; task version 1.0.1, Epoch	+1.3 pts
AIME 2025	85.2% AIME '25 (no tools)	85% Leaderboard, high	+0.2 pts

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-5 for developers (OpenAI) Aug 7, 2025

Its system card

Every result in the longer documents the lab published with it. 8 of them aren't in the launch post.