GPT-5.6 Luna

GPT-5.6 Luna places in the top three on 4 benchmarks.

OpenAI Released 9 Jul 2026

178

Benchmarks with a score

Ranked first, out of five or more models

In the top three

OpenAI Jul 2026

GPT-5.6 Luna

Closed weights

156.7 #30 of 271

Epoch Capabilities Index

MMMU-Pro 79.5% #13 /72
DeepSWE 67.2% #10 /28
Agents' Last Exam 5.74% #12 /12

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
Wer wird Millionär? LLM Benchmark	416667.0	#4 of 76	Official leaderboard
Creative Writing v3	78.5	#19 of 120	Official leaderboard
LMArena (Elo)	1,453	#59 of 333	Official leaderboard
Longform Creative Writing (EQ-Bench)	75.5	#27 of 119	Official leaderboard
FACTS Benchmark Suite	55.3%	#14 of 40	Official leaderboard
SimpleQA Verified	39.1%	#24 of 56	Official leaderboard
GraphWalks	81.3%	#9 of 21	Model card
SEA-HELM	72.1	#27 of 62	Official leaderboard
BullshitBench V2	41%	#55 of 122	Official leaderboard
Document Arena	1,462	#19 of 37	Official leaderboard
Extended NYT Connections	69.4%	#53 of 97	Official leaderboard
BullshitBench	25.4%	#61 of 107	Official leaderboard
FACTS Parametric V2	50.6%	#13 of 20	Official leaderboard
FACTS Parametric	25%	#31 of 45	Official leaderboard
Global MMLU Lite	83.1%	#30 of 43	Official leaderboard
FACTS Search V2	79.3%	#15 of 21	Official leaderboard
EQ-Bench 4	1,146	#23 of 32	Official leaderboard
Continual Learning Exam	25.3%	#9 of 10	Official leaderboard
OpenAI MRCR v2	41.3%	#9 of 9	Model card

Math and science

Benchmark	Score	Rank	Source
OTIS Mock AIME 2024-2025	98.3%	#26 of 202	Epoch AI
Verifying the Verifiers (Project APE CRED benchmark)	97.5%	#3 of 23	Official leaderboard
GPQA Diamond	91.6%	#29 of 219	Epoch AI
FrontierMath Tiers 1-3 v2	82.1%	#16 of 89	Epoch AI
CritPt	20.6%	#29 of 160	Epoch AI
ProofBench	54%	#11 of 59	Official leaderboard
FrontierMath Tier 4 v2	61%	#17 of 71	Epoch AI
DTBench	81.3%	#55 of 178	Official leaderboard
ARC-AGI	90.7%	#30 of 92	Official leaderboard
Surface Evolver Bench	61.9%	#9 of 26	Epoch AI
ARC-AGI-2	59.6%	#37 of 95	Official leaderboard
ProofBench v1.1	60%	#23 of 51	Official leaderboard
FrontierPhysics	3.60%	#11 of 19	Official leaderboard
SimpleBench	46.8%	#58 of 100	Official leaderboard
LatchBio Capabilities v1.0	33.5%	#24 of 27	Official leaderboard
MysteryMechanism	14.4%	#28 of 29	Official leaderboard
Terminal-Bench-Science	3.33%	#17 of 17	Official leaderboard

Coding

Benchmark	Score	Rank	Source
IOI (Vals)	72.9%	#4 of 55	Official leaderboard
ALE-Bench	1,667	#16 of 121	Epoch AI
AndroidBench	87.6%	#5 of 32	Official leaderboard
SWE-Bench Pro	62.7%	#20 of 77	Model card
WeirdML	60.9%	#30 of 112	Official leaderboard
Code Migration	44.5%	#22 of 78	Official leaderboard
Vibe Code Bench 1-100	22.6%	#8 of 26	Official leaderboard
Vibe Code Bench	77.1%	#36 of 115	Official leaderboard
DeepSWE	67.2%	#10 of 28	Official leaderboard
WebDev Arena	1,519	#38 of 105	Epoch AI
IOI v2 (Vals)	61.8%	#17 of 45	Official leaderboard
Code Arena	1,519	#51 of 124	Official leaderboard
SWE-sweep	2.50%	#5 of 10	Official leaderboard
Brokk Power Ranking	8.10	#10 of 19	Official leaderboard
FrontierCode	39.8%	#27 of 46	Epoch AI
Image-to-WebDev Arena	1,517	#35 of 55	Official leaderboard
SWE-rebench	43.6%	#9 of 13	Official leaderboard
WeirdML v3	5.16%	#22 of 25	Official leaderboard
CursorBench	35.9%	#17 of 19	Epoch AI
Vision2Web	64.5%	#10 of 10	Official leaderboard
SWE Refactor Bench	4.67%	#8 of 8	Official leaderboard
Artificial Analysis Coding Agent Index v1.1	74.6	#3 of 3	Model card

Agents and tools

Benchmark	Score	Rank	Source
PinchBench 2.0	88.7%	#7 of 57	Official leaderboard
LLM Chess	1,185	#16 of 124	Official leaderboard
LisanBench	4180.7	#27 of 133	Official leaderboard
BALROG	45.6%	#8 of 35	Official leaderboard
BrowseComp	83.3%	#23 of 77	Model card
Game Arena Poker	4.80	#10 of 31	Official leaderboard
MyPCBench	55.4%	#3 of 9	Official leaderboard
Chess Puzzles	43%	#30 of 83	Epoch AI
Toolathlon	53.4%	#16 of 37	Model card
GAIA	60.8%	#26 of 57	Official leaderboard
Mystery Game Puzzles	21%	#44 of 80	Epoch AI
Game Arena Chess	700	#23 of 39	Official leaderboard
Terminal-Bench 2.1	75.7%	#8 of 13	Official leaderboard
Game Arena Lines of Action	439	#23 of 36	Official leaderboard
SnakeBench	21	#121 of 183	Official leaderboard
Brood War Bench	50%	#4 of 6	Official leaderboard
Game Arena Word Art	259	#24 of 35	Official leaderboard
Game Arena Dark Hex	144	#27 of 38	Official leaderboard
Game Arena Bargaining	162	#26 of 36	Official leaderboard
Game Arena Word Association	92	#28 of 36	Official leaderboard
Terminal-Bench 4.0	17.3%	#18 of 23	Official leaderboard
Game Arena Four in a Row	189	#32 of 38	Official leaderboard
Game Arena Go	446	#31 of 36	Official leaderboard
Kaggle Game Arena	164	#34 of 39	Official leaderboard
ApprenticeBench	7.00%	#16 of 18	Official leaderboard
Game Arena Checkers	323	#34 of 38	Official leaderboard
OSWorld 2.0	45.6%	#17 of 19	Model card
Terminal-Bench 3.0	14.3%	#10 of 11	Official leaderboard
Game Arena Clobber	35	#33 of 36	Official leaderboard
Game Arena Nine Men's Morris	33	#34 of 36	Official leaderboard
Game Arena Ultimate Tic-Tac-Toe	21	#34 of 36	Official leaderboard
Game Arena Dots and Boxes	438	#35 of 37	Official leaderboard
ARC-AGI-3	0.18%	#23 of 24	Official leaderboard
AutomationBench	14.9%	#40 of 41	Model card
Game Arena Reversi	0	#36 of 36	Official leaderboard
Game Arena Coin Game	0	#35 of 35	Official leaderboard

Professional work

Benchmark	Score	Rank	Source
TaxEval v2	76.2%	#6 of 129	Official leaderboard
ProfBench	55.3%	#16 of 72	Official leaderboard
Excel Modeling Benchmark	67.1%	#18 of 76	Official leaderboard
PolicyBench	88.7%	#12 of 50	Official leaderboard
Vals Finance Agent v2	55%	#24 of 80	Official leaderboard
MortgageTax	67.3%	#26 of 86	Official leaderboard
LegalBench	84%	#43 of 138	Official leaderboard
CorpFin v2	64.2%	#40 of 119	Official leaderboard
MedScribe	84.4%	#36 of 98	Official leaderboard
GDPval-AA v2	1,592	#12 of 26	Model card
Legal Research Bench	36.5%	#43 of 88	Official leaderboard
BenchCAD	73.9%	#5 of 10	Model card
MedCode	42.4%	#49 of 96	Official leaderboard
Tax Agent Bench	24.8%	#43 of 79	Official leaderboard
HealthBench Consensus	95.1%	#10 of 17	Model card
Vending-Bench 2	$4,095	#40 of 66	Official leaderboard
HealthBench Hard	32%	#13 of 21	Model card
Public Benefits Bench	61.2%	#31 of 50	Official leaderboard
Harvey LAB (Legal Agent Benchmark)	1.25%	#56 of 80	Official leaderboard
APEX-Agents	43%	#42 of 56	Epoch AI
Vending-Bench Arena	$2,434	#24 of 32	Official leaderboard
HealthBench	55.8%	#24 of 30	Model card
CommerceAgentBench	44.9%	#11 of 13	Official leaderboard
HealthBench Professional	55.7%	#22 of 25	Model card
RCM-Bench-1	71.5%	#19 of 20	Official leaderboard
Agents' Last Exam	5.74%	#12 of 12	Official leaderboard
BigFinanceBench	36%	#2 of 2	Model card

Vision, audio and robotics

Benchmark	Score	Rank	Source
ZeroBench	21%	#10 of 66	Official leaderboard
MMMU-Pro	79.5%	#13 of 72	Model card
Vision Arena	1,257	#40 of 136	Official leaderboard
VoxelBench	1,854	#19 of 56	Official leaderboard
ParseBench	66.9%	#19 of 42	Official leaderboard
Furniture Assembly Benchmark	42.5%	#17 of 35	Epoch AI
GDP.pdf	22.7%	#21 of 39	Epoch AI
ArtifactArena	984	#15 of 21	Official leaderboard
Blueprint-Bench 2	22.6%	#27 of 35	Official leaderboard
Humanity's Sixth Sense	21.6%	#28 of 29	Official leaderboard

Safety and security

Benchmark	Score	Rank	Source
CyberBench (Vals)	83.9%	#2 of 26	Official leaderboard
Tacit Knowledge and Troubleshooting	83%	#4 of 29	Model card
CyberBench v1.1 (Vals)	73.6%	#10 of 43	Official leaderboard
DeepsecBench	26.6	#9 of 28	Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2)	62.5%	#23 of 71	Official leaderboard
DistressBench	72.7%	#17 of 42	Official leaderboard
TroubleshootingBench	37.7%	#8 of 19	Model card
Agent JSK	99.9%	#3 of 7	Model card
PCT (Polarized Contrastive Pairs)	44.7%	#23 of 46	Official leaderboard
BioSecBench-Surveillance	43.2%	#17 of 27	Official leaderboard
CVE-Bench	82.3%	#9 of 14	Model card
Multimodal Troubleshooting Virology	50.2%	#17 of 26	Model card
SpatialBench-Long	25%	#18 of 27	Official leaderboard
Capture the Flag (Internal)	93.2%	#2 of 3	Model card
Cyber Range	61.5%	#7 of 10	Model card
scBench-Long	13.1%	#19 of 27	Official leaderboard
Prompt injection robustness	97.1%	#5 of 7	Model card
ProtocolQA Open-Ended	29.9%	#18 of 25	Model card
SRE-Bench	1.53%	#24 of 33	Official leaderboard
TxBench-Preclinical-Pharmacology	43.7%	#21 of 28	Official leaderboard
EpiBench	19.8%	#21 of 27	Official leaderboard
First-Person Fairness evaluation	0.61%	#8 of 10	Model card
OncologyBench	10.7%	#25 of 30	Official leaderboard
ExploitBench	33.2%	#16 of 19	Model card
BioSecBench-Function	27.6%	#23 of 27	Official leaderboard
ExploitGym	12.4%	#12 of 14	Model card
VariantBench	27.1%	#24 of 27	Official leaderboard
MetagenomicsBench	24.3%	#24 of 26	Official leaderboard
TxBench-Antibody-Discovery	19.3%	#27 of 28	Official leaderboard
TxBench-Oligo-Discovery	23%	#27 of 28	Official leaderboard
SEC-bench Pro	48.9%	#10 of 10	Model card
Destructive action avoidance	73%	#8 of 8	Model card
Instruction hierarchy attacks	99.9%	#7 of 7	Model card
GeneBench Pro	10.8%	#4 of 4	Model card
LifeSciBench	51.2%	#4 of 4	Model card
Prompt injection attacks in search and function calling	89.7%	#3 of 3	Model card

Other

Benchmark	Score	Rank	Source
LMCA	48.5%	#29 of 139	Epoch AI
Vals Multimodal Index	69.1%	#7 of 33	Official leaderboard
Conceptual Reasoning Index	58.7	#41 of 149	Official leaderboard
Political Science LLM Benchmark	67%	#9 of 29	Official leaderboard
ACCoRD	66.6	#61 of 150	Official leaderboard
CAIS Vision Capabilities Index	53.7	#24 of 53	Official leaderboard
CAIS Risk Index	50.3	#31 of 65	Official leaderboard
CAIS Text Capabilities Index	31.3	#34 of 66	Official leaderboard
SAGE (Vals)	44.2%	#45 of 81	Official leaderboard
Artificial Analysis Intelligence Index v4.1	51.2	#6 of 7	Model card
Vals Science Index	25.3%	#22 of 23	Official leaderboard

OpenAI's models over time

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

Its launch post

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

GPT-5.6 (OpenAI) 9 Jul 2026

Agents' Last Exam 50.3
GDPval-AA v2 1591.8
Big Finance Bench 36
Artificial Analysis Intelligence Index v4.1 51.2
Artificial Analysis Coding Agent Index v1.1 74.6
SWE-Bench Pro 62.7
DeepSWE v1.1 67.2
Terminal-Bench 2.1 84.7
GeneBench Pro 10.8
LifeSciBench 51.2
HealthBench Professional 55.7
OSWorld 2.0 45.6
BrowseComp 83.3
BenchCAD (python tool) 73.9 (no tools 63.1) 73.9
SEC-Bench Pro 48.9
ExploitBench 33.2
ExploitGym 12.4
MMMU Pro (with tools) 79.5 (no tools 78.4) 79.5
gdp.pdf 22.7
GPQA Diamond 92.3
FrontierMath Tier 1-3 (v2) 78.6
FrontierMath Tier 4 (v2) 58.5
AutomationBench 14.9
Toolathlon 53.4
OpenAI MRCR v2 8-needle 256K-512K 41.3 (512K-1M 41.3) 41.3
GraphWalks BFS 256k f1 81.3 (1mil f1 51.2) 81.3
ARC-AGI-3 0.18

Its system cards

Every result in the longer documents the lab published with it. 30 of them aren't in the launch post.

GPT-5.6 Luna

GPT-5.6 Luna

Where it ranks

Knowledge and language

Math and science

Coding

Agents and tools

Professional work

Vision, audio and robotics

Safety and security

Other

OpenAI's models over time

Its launch post

Its system cards

Health

Hallucination

Safety

Robustness

Bias

Bio and chem risk

Cyber risk

AI R&D risk

Health

Hallucination

Safety

Robustness

Bias

Bio and chem risk

Cyber risk

AI R&D risk

Health

Hallucination

Safety

Robustness

Bio and chem risk

Cyber risk