GPT-5.6 Terra

GPT-5.6 Terra places in the top three on 5 benchmarks.

OpenAI Released 9 Jul 2026

164

Benchmarks with a score

Ranked first, out of five or more models

In the top three

OpenAI Jul 2026

GPT-5.6 Terra

Closed weights

160.1 #15 of 271

Epoch Capabilities Index

MMMU-Pro 82% #5 /72
DeepSWE 69.6% #6 /28
Terminal-Bench 4.0 21.5% #15 /23

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
LMArena (Elo)	1,467	#40 of 333	Official leaderboard
Creative Writing v3	79.5	#16 of 120	Official leaderboard
Wer wird Millionär? LLM Benchmark	183333.0	#13 of 76	Official leaderboard
Longform Creative Writing (EQ-Bench)	78.0	#22 of 119	Official leaderboard
SEA-HELM	75.5	#12 of 62	Official leaderboard
OpenAI MRCR v2	89.6%	#2 of 9	Model card
Document Arena	1,477	#11 of 37	Official leaderboard
BullshitBench V2	54%	#40 of 122	Official leaderboard
Global MMLU Lite	92.7%	#15 of 43	Official leaderboard
FACTS Parametric	56.7%	#17 of 45	Official leaderboard
SimpleQA Verified	41.5%	#22 of 56	Official leaderboard
BullshitBench	34.5%	#44 of 107	Official leaderboard
FACTS Benchmark Suite	50%	#17 of 40	Official leaderboard
FACTS Search V2	81.7%	#9 of 21	Official leaderboard
Extended NYT Connections	78.4%	#42 of 97	Official leaderboard
EQ-Bench 4	1,220	#15 of 32	Official leaderboard
FACTS Parametric V2	57.6%	#10 of 20	Official leaderboard
Continual Learning Exam	33.2%	#5 of 10	Official leaderboard
GraphWalks	76.9%	#11 of 21	Model card
FACTS Grounding v2	60.5%	#11 of 15	Official leaderboard

Math and science

Benchmark	Score	Rank	Source
ProofBench	71%	#4 of 59	Official leaderboard
CritPt	30%	#11 of 160	Epoch AI
OTIS Mock AIME 2024-2025	99.7%	#14 of 202	Epoch AI
GPQA Diamond	93.3%	#18 of 219	Epoch AI
FrontierMath Tiers 1-3 v2	86%	#11 of 89	Epoch AI
ARC-AGI	96.5%	#12 of 92	Official leaderboard
ARC-AGI-2	83.9%	#15 of 95	Official leaderboard
Verifying the Verifiers (Project APE CRED benchmark)	97%	#4 of 23	Official leaderboard
DTBench	88.9%	#32 of 178	Official leaderboard
FrontierMath Tier 4 v2	70.7%	#15 of 71	Epoch AI
Surface Evolver Bench	83.8%	#6 of 26	Epoch AI
ProofBench v1.1	74%	#18 of 51	Official leaderboard
SimpleBench	48.9%	#55 of 100	Official leaderboard
FrontierPhysics	3.00%	#12 of 19	Official leaderboard
Terminal-Bench-Science	8.57%	#11 of 17	Official leaderboard
Human Behavior Bench	77.8	#8 of 11	Official leaderboard

Coding

Benchmark	Score	Rank	Source
ALE-Bench	1,951	#11 of 121	Epoch AI
WeirdML	78.3%	#11 of 112	Official leaderboard
IOI (Vals)	65.3%	#7 of 55	Official leaderboard
IOI v2 (Vals)	87.6%	#7 of 45	Official leaderboard
Code Migration	47.8%	#13 of 78	Official leaderboard
DeepSWE	69.6%	#6 of 28	Official leaderboard
AndroidBench	86.8%	#7 of 32	Official leaderboard
SWE-Bench Pro	63.4%	#17 of 77	Model card
WebDev Arena	1,521	#36 of 105	Epoch AI
Vibe Code Bench	74.6%	#40 of 115	Official leaderboard
Code Arena	1,520	#49 of 124	Official leaderboard
Senior SWE-Bench	27.4%	#11 of 23	Official leaderboard
SWE-Bench Pro V2 Hard	86.3%	#8 of 16	Official leaderboard
FrontierCode	41.3%	#24 of 46	Epoch AI
Image-to-WebDev Arena	1,541	#29 of 55	Official leaderboard
MirrorCode	41.1%	#15 of 28	Epoch AI
SWE-sweep	1.50%	#6 of 10	Official leaderboard
Artificial Analysis Coding Agent Index v1.1	77.4	#2 of 3	Model card
Vibe Code Bench 1-100	14.8%	#19 of 26	Official leaderboard
CursorBench	41.3%	#14 of 19	Epoch AI
SWE-Bench Pro V2	92.4%	#14 of 16	Official leaderboard
Vision2Web	67.5%	#9 of 10	Official leaderboard

Agents and tools

Benchmark	Score	Rank	Source
LLM Chess	1,293	#12 of 124	Official leaderboard
BrowseComp	87.5%	#13 of 77	Model card
BALROG	53.2%	#6 of 35	Official leaderboard
SnakeBench	30	#38 of 183	Official leaderboard
Game Arena Poker	7.27	#7 of 31	Official leaderboard
LisanBench	3302.7	#34 of 133	Official leaderboard
Chess Puzzles	47%	#23 of 83	Epoch AI
Mystery Game Puzzles	35%	#23 of 80	Epoch AI
GAIA	71.8%	#17 of 57	Official leaderboard
PinchBench 2.0	75.9%	#22 of 57	Official leaderboard
Toolathlon	53.1%	#17 of 37	Model card
Terminal-Bench 2.1	78.4%	#6 of 13	Official leaderboard
Game Arena Chess	725	#22 of 39	Official leaderboard
ARC-AGI-3	0.80%	#14 of 24	Official leaderboard
ApprenticeBench	16%	#11 of 18	Official leaderboard
Agent Arena	-0.01	#33 of 53	Official leaderboard
Game Arena Word Art	295	#22 of 35	Official leaderboard
Terminal-Bench 3.0	20.8%	#7 of 11	Official leaderboard
Game Arena Word Association	175	#23 of 36	Official leaderboard
Terminal-Bench 4.0	21.5%	#15 of 23	Official leaderboard
Game Arena Checkers	744	#25 of 38	Official leaderboard
Game Arena Lines of Action	433	#24 of 36	Official leaderboard
Game Arena Clobber	154	#25 of 36	Official leaderboard
Game Arena Go	512	#26 of 36	Official leaderboard
OSWorld 2.0	50.2%	#14 of 19	Model card
Kaggle Game Arena	195	#29 of 39	Official leaderboard
Game Arena Dark Hex	131	#29 of 38	Official leaderboard
Game Arena Dots and Boxes	621	#29 of 37	Official leaderboard
Game Arena Ultimate Tic-Tac-Toe	57	#30 of 36	Official leaderboard
Brood War Bench	50%	#5 of 6	Official leaderboard
Game Arena Coin Game	200	#30 of 35	Official leaderboard
Game Arena Bargaining	136	#31 of 36	Official leaderboard
Game Arena Reversi	151	#31 of 36	Official leaderboard
Game Arena Nine Men's Morris	260	#32 of 36	Official leaderboard
Game Arena Four in a Row	91	#36 of 38	Official leaderboard
AutomationBench	15.2%	#39 of 41	Model card

Professional work

Benchmark	Score	Rank	Source
TaxEval v2	76.2%	#7 of 129	Official leaderboard
ProfBench	57.2%	#11 of 72	Official leaderboard
LegalBench	85.1%	#25 of 138	Official leaderboard
Vending-Bench Arena	$9,191	#6 of 32	Official leaderboard
Agents' Last Exam	50.4%	#5 of 25	Model card
CorpFin v2	65.3%	#32 of 119	Official leaderboard
Vending-Bench 2	$7,343	#19 of 66	Official leaderboard
MortgageTax	67.3%	#25 of 86	Official leaderboard
BenchCAD	78.2%	#3 of 10	Model card
Excel Modeling Benchmark	66.2%	#23 of 76	Official leaderboard
APEX-Agents	58.2%	#17 of 56	Epoch AI
Legal Research Bench	41.4%	#28 of 88	Official leaderboard
Quant-Arena	47.6	#3 of 9	Official leaderboard
Vals Finance Agent v2	54.4%	#27 of 80	Official leaderboard
PolicyBench	87.6%	#17 of 50	Official leaderboard
Tax Agent Bench	30%	#30 of 79	Official leaderboard
GDPval-AA v2	1,593	#11 of 26	Model card
MedCode	43.4%	#43 of 96	Official leaderboard
MedScribe	82.9%	#47 of 98	Official leaderboard
BigFinanceBench	51%	#1 of 2	Model card
Public Benefits Bench	62.4%	#28 of 50	Official leaderboard
HealthBench Hard	32.7%	#12 of 21	Model card
HealthBench Consensus	95.1%	#11 of 17	Model card
HealthBench	57%	#22 of 30	Model card
Harvey LAB (Legal Agent Benchmark)	0.83%	#60 of 80	Official leaderboard
RCM-Bench-1	87%	#15 of 20	Official leaderboard
HealthBench Professional	57.7%	#19 of 25	Model card
Parametric CAD Bench	17.2%	#11 of 11	Official leaderboard

Vision, audio and robotics

Benchmark	Score	Rank	Source
MMMU-Pro	82%	#5 of 72	Model card
ZeroBench	19%	#14 of 66	Official leaderboard
Vision Arena	1,268	#30 of 136	Official leaderboard
VoxelBench	1,895	#16 of 56	Official leaderboard
Furniture Assembly Benchmark	54.2%	#11 of 35	Epoch AI
GDP.pdf	24.7%	#13 of 39	Epoch AI
ParseBench	67.1%	#18 of 42	Official leaderboard
ArtifactArena	1,093	#10 of 21	Official leaderboard
Blueprint-Bench 2	30.8%	#19 of 35	Official leaderboard
1H-VideoQA	67.3%	#15 of 20	Official leaderboard
Humanity's Sixth Sense	22.7%	#27 of 29	Official leaderboard

Safety and security

Benchmark	Score	Rank	Source
Human Pathogen Capabilities Test v2 (HPCT v2)	70%	#6 of 71	Official leaderboard
Tacit Knowledge and Troubleshooting	84.1%	#3 of 29	Model card
DistressBench	76.4%	#10 of 42	Official leaderboard
Agent JSK	100%	#2 of 7	Model card
Prompt injection attacks in search and function calling	94.6%	#1 of 3	Model card
CyberBench v1.1 (Vals)	72.1%	#15 of 43	Official leaderboard
SRE-Bench	9.54%	#12 of 33	Official leaderboard
ExploitGym	23.2%	#6 of 14	Model card
DeepsecBench	19.1	#14 of 28	Official leaderboard
PCT (Polarized Contrastive Pairs)	46.2%	#25 of 46	Official leaderboard
First-Person Fairness evaluation	0.88%	#6 of 10	Model card
Destructive action avoidance	81%	#5 of 8	Model card
ExploitBench	52.9%	#12 of 19	Model card
GeneBench Pro	23.3%	#3 of 4	Model card
LifeSciBench	56%	#3 of 4	Model card
SEC-bench Pro	57.7%	#8 of 10	Model card
Instruction hierarchy attacks	99.9%	#6 of 7	Model card
Prompt injection robustness	96.7%	#6 of 7	Model card
Capture the Flag (Internal)	91.8%	#3 of 3	Model card

Other

Benchmark	Score	Rank	Source
LMCA	55%	#15 of 139	Epoch AI
Conceptual Reasoning Index	65.7	#18 of 149	Official leaderboard
Political Science LLM Benchmark	68.3%	#5 of 29	Official leaderboard
ACCoRD	74.7	#27 of 150	Official leaderboard
CAIS Text Capabilities Index	38.2	#21 of 66	Official leaderboard
SAGE v1.1 (Vals)	87%	#10 of 28	Official leaderboard
CAIS Risk Index	46.7	#24 of 65	Official leaderboard
SAGE (Vals)	47%	#33 of 81	Official leaderboard
Vals Multimodal Index	65.1%	#14 of 33	Official leaderboard
CAIS Vision Capabilities Index	55.1	#23 of 53	Official leaderboard
LiveBench	67.8%	#14 of 28	Official leaderboard
Artificial Analysis Intelligence Index v4.1	55.0	#5 of 7	Model card

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 50.4
GDPval-AA v2 1593
Big Finance Bench 51
Artificial Analysis Intelligence Index v4.1 55
Artificial Analysis Coding Agent Index v1.1 77.4
SWE-Bench Pro 63.4
DeepSWE v1.1 69.6
Terminal-Bench 2.1 87.4
GeneBench Pro 23.3
LifeSciBench 56
HealthBench Professional 57.7
OSWorld 2.0 50.2
BrowseComp 87.5
BenchCAD (python tool) 78.2 (no tools 62.3) 78.2
SEC-Bench Pro 57.7
ExploitBench 52.9
ExploitGym 23.2
MMMU Pro (with tools) 82 (no tools 80.7) 82
gdp.pdf 24.7
GPQA Diamond 92.9
FrontierMath Tier 1-3 (v2) 84.9
FrontierMath Tier 4 (v2) 68.3
AutomationBench 15.2
Toolathlon 53.1
OpenAI MRCR v2 8-needle 256K-512K 89.6 (512K-1M 72.5) 89.6
GraphWalks BFS 256k f1 76.9 (1mil f1 71.2) 76.9
ARC-AGI-3 0.8

Its system cards

Every result in the longer documents the lab published with it. 31 of them aren't in the launch post.

GPT-5.6 Terra

GPT-5.6 Terra

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

Monitorability

Bio and chem risk

Cyber risk

AI R&D risk

Health

Hallucination

Safety

Robustness

Bias

Monitorability

Bio and chem risk

Cyber risk

AI R&D risk