Benchmark Almanac

GPT-5.4

GPT-5.4 ranks first on 21 of the 327 benchmarks with five or more models scored.

OpenAI Released March 5th, 2026 Family GPT-5 The lab's model card Compare with others
336
Benchmarks with a score
21
Ranked first, out of five or more models
72
In the top three
OpenAI Mar 2026

GPT-5.4

Closed weights Flagship

157.8 #27 of 256
Epoch Capabilities Index
Self-Reported Index 158.2 #42 of 404
  • Humanity's Last Exam 36.2% #9 /47
  • SWE-bench Verified 76.9% #8 /32
  • SWE-Bench Pro 59.1% #2 /24

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

BenchmarkScoreRankSource
SnorkelSpatial 99% #1 of 27 Official leaderboard
CL-bench 27.9% #1 of 24 Official leaderboard
MMLU-Pro 87.5% #9 of 212 Official leaderboard
Structured Output Benchmark 87% #1 of 19 Official leaderboard
Global PIQA 91.3% #3 of 44 Official leaderboard
HELM Arabic Enterprise 76.7 #2 of 23 Official leaderboard
LMArena (Elo) 1,475 #30 of 335 Official leaderboard
SkateBench v2 81.5% #2 of 21 Official leaderboard
HELM Arabic 82.5 #3 of 28 Official leaderboard
Wer wird Millionär? LLM Benchmark 275000.0 #9 of 77 Official leaderboard
CL-bench Life 21.7% #2 of 16 Official leaderboard
KINA 48.5% #5 of 37 Official leaderboard
SEA-HELM 76.0 #10 of 71 Official leaderboard
Creative Writing v3 78.8 #18 of 120 Official leaderboard
Longform Creative Writing (EQ-Bench) 78.3 #19 of 119 Official leaderboard
Judgemark v2.1 85.0 #7 of 42 Official leaderboard
EQ-Bench 3 1,537 #14 of 78 Official leaderboard
MultiNRC 58.3% #7 of 38 Official leaderboard
GraphWalks 93% #5 of 27 Lab model card
Humanity's Last Exam 36.2% #9 of 47 Official leaderboard
HLE Extended 56.1% #6 of 31 Official leaderboard
Extended NYT Connections 91.3% #20 of 103 Official leaderboard
Lech Mazur Creative Writing 1.92 #11 of 50 Official leaderboard
HUMAINE 30 #14 of 50 Official leaderboard
EQ-Bench 4 1,258 #10 of 32 Official leaderboard
Vectara Hallucination Leaderboard (2025 dataset) 7.00% #33 of 104 Official leaderboard
OpenAI MRCR v2 52.7% #15 of 47 Official leaderboard
AA-LCR Extended 68.3% #17 of 49 Official leaderboard
Judgemark v4 72.1 #14 of 40 Official leaderboard
Document Arena 1,471 #14 of 39 Official leaderboard
Continual Learning Bench 0.19 #2 of 5 Official leaderboard
DesignArena ASCII Art 1,194 #31 of 77 Official leaderboard
BullshitBench V2 48% #55 of 129 Official leaderboard
STALE 15.7% #5 of 10 Official leaderboard
BullshitBench 25.4% #59 of 107 Official leaderboard
SimpleQA Verified 30.4% #31 of 56 Official leaderboard
Hemingway-bench 985 #31 of 42 Official leaderboard
Joke Arena 1,477 #20 of 25 Official leaderboard
Corp Speak Arena 1,364 #21 of 22 Official leaderboard

Math and science

BenchmarkScoreRankSource
FrontierMath 47.6% #2 of 55 Official leaderboard
SnorkelGraph 84.5% #1 of 24 Official leaderboard
HMMT 2026 97.7% #2 of 31 Official leaderboard
GPQA Diamond 93.3% #22 of 229 Epoch AI run
AIME 2026 99.2% #3 of 31 Official leaderboard
IMProofBench 49.3% #1 of 9 Official leaderboard
LiveMathematicianBench 45.2% #2 of 17 Official leaderboard
FrontierMath Tier 4 27.1% #7 of 55 Epoch AI run
MathArena (Apex, IMO, USAMO) 54.2% #6 of 47 Official leaderboard
SpatialBench (LatchBio) 57.4% #2 of 14 Official leaderboard
ProofBench 56% #9 of 59 Official leaderboard
SnorkelSequences 71.6% #4 of 25 Official leaderboard
DTBench 94.4% #29 of 180 Official leaderboard
OTIS Mock AIME 2024-2025 97.8% #37 of 212 Epoch AI run
KellyBench -7.90% #1 of 5 Official leaderboard
USAMO 2026 95.2% #2 of 9 Official leaderboard
MathArena Apex Shortlist 81.4% #9 of 38 Official leaderboard
FrontierMath Tiers 1-3 v2 78.6% #21 of 86 Official leaderboard
ARC-AGI-2 74% #24 of 98 Official leaderboard
ARC-AGI 93.7% #26 of 95 Official leaderboard
FrontierMath Tier 4 v2 49% #21 of 68 Official leaderboard
ResearchClawBench 18.8 #10 of 23 Official leaderboard
AstaBench 46.5% #6 of 13 Official leaderboard
Project Euler (MathArena) 89% #2 of 4 Official leaderboard
Leipzig Benchmark 28% #11 of 19 Official leaderboard
xbench-ScienceQA 70.8% #17 of 29 Official leaderboard
Echo General AI Prediction Leaderboard 999 #10 of 17 Official leaderboard
ForecastBench 0.14 #49 of 82 Official leaderboard
FrontierScience-Research 33% #3 of 5 Lab model card
NatureBench 6.67% #14 of 18 Official leaderboard
BTF-2 0.15 #3 of 3 Official leaderboard

Coding

BenchmarkScoreRankSource
SWE-Bench Pro 59.1% #2 of 24 Official leaderboard
IOI (Vals) 67.8% #6 of 55 Official leaderboard
WeirdML 77.7% #14 of 112 Official leaderboard
AlgoTune 1.85× #3 of 18 Official leaderboard
OpenHands Index 64.3% #6 of 33 Official leaderboard
QwenVision2Code 1,884 #1 of 5 Official leaderboard
LiveSQLBench 31% #8 of 36 Official leaderboard
METR Time Horizon 5.7 h #6 of 26 Claim in metr.org, via Epoch AI
METR Time Horizon 1.1 5.7 h #6 of 26 Official leaderboard
SWE-bench Verified 76.9% #8 of 32 Epoch AI run
MLE-bench 30 23.3% #2 of 8 Lab model card
PostTrainBench 28.2% #6 of 23 Official leaderboard
RoadmapBench 29.6% #3 of 11 Official leaderboard
ALE-Bench 1,811 #35 of 126 Official leaderboard
FeatureBench 66.2% #5 of 16 Official leaderboard
FeatureBench Lite 66.2% #5 of 16 Official leaderboard
GSO-Bench 31.4% #10 of 30 Official leaderboard
SWE-Bench Pro Private 43.4% #4 of 12 Official leaderboard
SWE Atlas Test Writing 44.4% #8 of 22 Official leaderboard
DesignArena Game Dev 1,229 #53 of 143 Official leaderboard
AndroidBench 74.1% #12 of 32 Official leaderboard
SlopCodeBench 25.5% #9 of 22 Official leaderboard
FrogsGame post-training task 2.00% #7 of 17 Official leaderboard
DesignArena Data Visualization 1,233 #49 of 118 Official leaderboard
DesignArena UI Component 1,231 #59 of 139 Official leaderboard
ts-bench 60% #3 of 7 Official leaderboard
DesignArena SVG 1,185 #45 of 103 Official leaderboard
Vibe Code Bench 67.4% #53 of 120 Official leaderboard
Vision2Web S2-2026 37.1% #4 of 9 Official leaderboard
SWE Atlas 40.8% #10 of 22 Official leaderboard
SWE Atlas Refactoring 44.3% #8 of 17 Official leaderboard
Code Migration 35% #37 of 78 Official leaderboard
GBAEval 45.1% #10 of 21 Official leaderboard
WebDev Arena 1,463 #52 of 105 arena.ai, via Epoch AI
DesignArena Website 1,208 #75 of 151 Official leaderboard
DesignArena Code 1,200 #73 of 143 Official leaderboard
Code Arena 1,462 #67 of 128 Official leaderboard
OpenAI-Proof Q&A 4.16% #5 of 9 Lab model card
FrontierSWE v1 46% #10 of 17 Official leaderboard
Monorepo-Bench 59.3% #3 of 5 Lab model card
SWE Atlas Codebase QnA Extended 10.7% #23 of 38 Official leaderboard
GameDevBench 52% #11 of 17 Official leaderboard
TasteVal 21.3× #13 of 20 Official leaderboard
SWE-bench Verified Extended 40% #29 of 44 Official leaderboard
ProgramBench 0% #17 of 25 Official leaderboard
AutoLab 0.36 #7 of 10 Official leaderboard
DesignArena 3D Design 1,084 #98 of 136 Official leaderboard
APEX-SWE 34.3% #47 of 65 Official leaderboard
DeepSWE 51.8% #22 of 28 Official leaderboard
ProgramBench tests passed 37.7% #20 of 25 Official leaderboard
Image-to-WebDev Arena 1,441 #45 of 55 Official leaderboard
MirrorCode 23.3% #33 of 39 Epoch AI run
DesignArena Agentic Overall 1,034 #64 of 74 Official leaderboard
Senior SWE-Bench 16.8% #20 of 23 Official leaderboard
DesignArena Full Stack 996 #45 of 49 Official leaderboard
Vals RSI Index 0.15 #25 of 27 Official leaderboard
DesignArena React Native 1,081 #52 of 56 Official leaderboard
DesignArena Frontend Web App 1,040 #50 of 53 Official leaderboard
DesignArena Android (Kotlin) 1,021 #44 of 46 Official leaderboard
ProgramBench almost solved 0% #24 of 25 Official leaderboard
PostTrainBench 1.2 17.8% #21 of 21 Official leaderboard
PostTrainBench 1.1 19% #11 of 11 Official leaderboard
NanoGPT-Bench 8.60% #2 of 2 Official leaderboard

Agents and tools

BenchmarkScoreRankSource
tau2-bench 98.9% #1 of 79 Lab model card
Pencil Puzzle Bench 70.2% #3 of 53 Official leaderboard
Toolathlon 54.6% #3 of 50 Official leaderboard
LLM Chess 1,322 #11 of 123 Official leaderboard
GAIA 84.7% #7 of 64 Official leaderboard
Game Arena Werewolf 0 #1 of 8 Official leaderboard
OpenComputer 68.3% #1 of 7 Official leaderboard
WebTailBench 57.4% #1 of 7 Official leaderboard
Game Arena Dark Hex 387 #6 of 41 Official leaderboard
Game Arena Five in a Row 878 #3 of 20 Official leaderboard
ClawProBench 67.3% #8 of 49 Official leaderboard
Claw-Eval-Live 63.8% #2 of 12 Official leaderboard
clembench 88.8 #1 of 6 Official leaderboard
CocoaBench 45.1% #1 of 6 Official leaderboard
LisanBench 5334.7 #23 of 133 Official leaderboard
DeepSearchQA 63.7% #2 of 10 Official leaderboard
Online-Mind2Web 92.7% #1 of 5 Official leaderboard
Claw-Eval 78.4% #5 of 24 Official leaderboard
Game Arena Checkers 1,095 #9 of 41 Official leaderboard
Chess Puzzles 48% #21 of 93 Epoch AI run
WildClawBench 56.8% #7 of 31 Official leaderboard
AutoResearchBench 7.44% #3 of 13 Official leaderboard
SnakeBench 29 #47 of 188 Official leaderboard
CUA-World-Long 27.5% #1 of 4 Official leaderboard
STATE-Bench Main 0.7.0 55.7% #1 of 4 Official leaderboard
Game Arena Reversi 756 #10 of 39 Official leaderboard
Mystery Game Puzzles 37% #22 of 84 Epoch AI run
Game Arena Clobber 449 #11 of 39 Official leaderboard
WebArena-Verified 67.3% #2 of 7 Lab model card
UniClawBench 40.8% #3 of 10 Official leaderboard
Clocktower Radio 1,661 #7 of 23 Official leaderboard
BrowseComp 82.7% #27 of 81 Lab model card
Game Arena Coin Game 416 #13 of 39 Official leaderboard
Game Arena Lines of Action 489 #13 of 39 Official leaderboard
MM-ClawBench 73.6% #2 of 6 Official leaderboard
Kaggle Game Arena 353 #15 of 42 Official leaderboard
Cresta-Bench 65.1% #5 of 14 Official leaderboard
Game Arena Dots and Boxes 941 #15 of 40 Official leaderboard
Claw-Eval Pass^3 60.3% #9 of 24 Official leaderboard
OSWorld-Verified 75% #16 of 41 Lab model card
τ³-bench Banking 39.4% #10 of 25 Official leaderboard
PinchBench 2.0 75.7% #23 of 57 Official leaderboard
Game Arena Ultimate Tic-Tac-Toe 366 #16 of 39 Official leaderboard
QwenClawBench 56.7% #5 of 12 Official leaderboard
Game Arena Four in a Row 434 #18 of 43 Official leaderboard
Game Arena Poker 1.50 #13 of 31 Official leaderboard
Game Arena Chess 894 #18 of 42 Official leaderboard
ADeptS-Bench Desktop 66.1% #3 of 7 Official leaderboard
ADeptS-Bench Mobile 62.7% #3 of 7 Official leaderboard
EnvSimBench 42% #3 of 7 Official leaderboard
Gaia2-CLI 55.6% #3 of 7 Official leaderboard
PlanBench-XL 51.9% #4 of 9 Official leaderboard
WeaveBench 22.8% #4 of 9 Official leaderboard
Game Arena Word Association 282 #19 of 40 Official leaderboard
Game Arena Word Art 605 #19 of 39 Official leaderboard
iOSWorld 39.8% #3 of 6 Official leaderboard
RuneBench 4.70 #40 of 76 Official leaderboard
Game Arena Bargaining 186 #21 of 39 Official leaderboard
Game Arena Nine Men's Morris 610 #21 of 39 Official leaderboard
LMArena Search Arena 1,196 #20 of 37 Official leaderboard
Terminal-Bench 2.1 Extended 31.3% #24 of 44 Official leaderboard
EBR-bench 25.4% #18 of 31 Epoch AI run
Game Arena Go 856 #23 of 39 Official leaderboard
DuelLab GameBench 2 62.1 #34 of 57 Official leaderboard
EdgeBench 39.3 #3 of 5 Official leaderboard
Toloka Arena 24.9% #3 of 5 Official leaderboard
MCP Atlas 70.6% #23 of 37 Official leaderboard
AutomationBench 11.6% #27 of 42 Official leaderboard
ClawBench 25.5% #4 of 6 Official leaderboard
CAR-bench 50% #9 of 12 Official leaderboard
ApprenticeBench 11% #14 of 18 Official leaderboard
BrowseComp Extended 14% #33 of 42 Official leaderboard
PinchBench 59.9% #38 of 48 Official leaderboard
Clocktower Bench 1,276 #9 of 11 Official leaderboard
Long-Horizon-Terminal-Bench (LHTB) 27.2% #21 of 25 Official leaderboard
ARC-AGI-3 0.21% #24 of 27 Official leaderboard
DeepResearch Bench 35.1% #22 of 22 evals.futuresearch.ai, via Epoch AI
Deep Research Bench (FutureSearch) 0.35 #21 of 21 Official leaderboard
STATE-Bench 58.6% #2 of 2 Official leaderboard

Professional work

BenchmarkScoreRankSource
SnorkelUnderwrite 91% #1 of 27 Official leaderboard
HealthBench Consensus 96.6% #1 of 17 Lab model card
TaxCalcBench 62.8% #1 of 17 Official leaderboard
Vals Index v1 66.1% #3 of 34 Official leaderboard
Finance Reasoning (Snorkel) 52% #2 of 21 Official leaderboard
LegalBench 86% #17 of 138 Official leaderboard
ProfBench 56.8% #12 of 72 Official leaderboard
HealthBench Professional 59% #1 of 6 Official leaderboard
MortgageTax 68.3% #16 of 87 Official leaderboard
prinzbench 69.0 #5 of 26 Official leaderboard
Vals Finance Agent 57.1% #11 of 49 Official leaderboard
HealthBench 62.6% #7 of 31 Lab model card
MedXpertQA MM Extended 62.1% #10 of 41 Official leaderboard
CorpFin v2 65.3% #33 of 119 Official leaderboard
GDPval 83% #3 of 10 Lab model card
CaseLaw v2 63.8% #16 of 53 Official leaderboard
TaxEval v2 74% #43 of 129 Official leaderboard
HealthBench Hard 40.1% #7 of 21 Lab model card
BigLaw Bench 91% #1 of 3 Lab model card
Vending-Bench 2 $6,144 #25 of 67 Official leaderboard
PhysicianBench 27.7% #5 of 13 Official leaderboard
Investment Banking Modeling Tasks (OpenAI) 87.3% #2 of 5 Lab model card
APEX-1 65.6% #29 of 71 Official leaderboard
3DCodeArena 1,042 #6 of 14 Official leaderboard
APEX-Agents 1.1 68.2% #29 of 65 Official leaderboard
GDPval Extended 1.10% #20 of 43 Official leaderboard
TRLawBench 9.08 #17 of 34 Official leaderboard
BankerToolBench v1.1 13.7% #5 of 10 Official leaderboard
PatientAgentBench 4.16 #4 of 8 Official leaderboard
APEX-Agents 52.4% #31 of 61 Collected by Epoch AI, source not named
MedCode 41.3% #52 of 96 Official leaderboard
QF-Bench 57.5% #5 of 9 Official leaderboard
PRBench 44.4% #27 of 48 Official leaderboard
FORTE 68.8% #12 of 21 Official leaderboard
Agents' Last Exam 20.5% #18 of 31 Official leaderboard
YC-Bench $1,000,803 #35 of 59 Official leaderboard
CHI-Bench 16% #14 of 23 Official leaderboard
MedScribe 77.5% #63 of 97 Official leaderboard
OfficeQA Pro 41.4% #8 of 12 Official leaderboard
Workspace-Bench-Lite 51.8% #6 of 9 Official leaderboard
MBABench 16.0 #4 of 6 Official leaderboard
EnterpriseBench: CoreCraft 36.4% #36 of 51 Official leaderboard
OfficeQA 68.1% #10 of 14 Lab model card
Vending-Bench Arena $2,158 #25 of 32 Official leaderboard
Harvey LAB (Legal Agent Benchmark) 0% #69 of 80 Official leaderboard

Vision, audio and robotics

BenchmarkScoreRankSource
MATH-Vision (MathVision) 96.1% #1 of 105 Official leaderboard
AgentWorldBench 58.3 #1 of 13 Official leaderboard
Math Kangaroo 2025 (MathArena) 92.5% #2 of 22 Official leaderboard
MeasureBench Synthetic 38.4% #2 of 19 Official leaderboard
ZeroBench 41% #10 of 76 Official leaderboard
Vision Arena 1,282 #19 of 136 Official leaderboard
VISTA 50.9% #8 of 54 Official leaderboard
MMMU-Pro 81.2% #14 of 90 Lab model card
MeasureBench Real-world 32.9% #3 of 19 Official leaderboard
Claw-Eval-MM 54.4% #2 of 11 Official leaderboard
CiteVQA 59% #3 of 16 Official leaderboard
ArtifactArena 1,179 #5 of 21 Official leaderboard
ClockBench 50.6% #10 of 37 Official leaderboard
MMMU-Pro Extended 39.8% #11 of 38 Official leaderboard
CharXiv Extended 82.7% #12 of 38 Official leaderboard
vCX Benchmark 89.6% #14 of 42 Official leaderboard
Chartography 29.6% #16 of 43 Official leaderboard
MineBench 1,651 #28 of 75 Official leaderboard
WorldBench 58.2% #3 of 8 Official leaderboard
VoxelBench Image 1,421 #5 of 13 Official leaderboard
OVO-S-Bench 50.9% #6 of 15 Official leaderboard
BabyVision 49.7% #14 of 27 Official leaderboard
ParseBench 61.7% #26 of 44 Official leaderboard
Furniture Assembly Benchmark 37.5% #21 of 35 Official leaderboard
SpatialWorld Physical 6.60% #6 of 9 Official leaderboard
Blueprint-Bench 2 27.1% #26 of 38 Official leaderboard
VoxelBench 1,424 #43 of 56 Official leaderboard
BVB 60.5 #35 of 45 Official leaderboard
SpatialWorld Digital 11.9% #7 of 9 Official leaderboard
CC-OCR-Bench-V2 56.7% #13 of 15 Official leaderboard

Safety and security

BenchmarkScoreRankSource
Capture the Flag (Professional) 88.2% #3 of 21 Lab model card
SMDD-Bench 40.2% #1 of 7 Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2) 65.3% #12 of 71 Official leaderboard
nocot-bench (NCRI) 94.0 #33 of 185 Official leaderboard
ProtocolQA Open-Ended 42.5% #5 of 27 Lab model card
KORA 58.8% #12 of 42 Official leaderboard
SEC-bench Pro 31.7% #2 of 6 Official leaderboard
SEC-bench Pro 260617 39% #2 of 6 Official leaderboard
BioTIER Refuse 76.9% #24 of 69 Official leaderboard
Destructive action avoidance 86% #3 of 8 Lab model card
CyberGym 79% #13 of 34 Official leaderboard
CyberBench (Vals) 72.9% #10 of 26 Official leaderboard
EvalDetectBench 0.82 #4 of 10 Official leaderboard
CyberGym-E2E 65.9% #2 of 5 Official leaderboard
GeneBench Pro 16.3% #5 of 12 Official leaderboard
HarnessAudit-Bench 0.32 #3 of 7 Official leaderboard
ABC Bench Fragment Design 90.9% #12 of 27 Official leaderboard
Dictatorship Eval 75.7% #10 of 20 Official leaderboard
TroubleshootingBench 35.8% #10 of 20 Lab model card
CVE-Bench 86.3% #7 of 14 Lab model card
Cyber Range 73.3% #5 of 10 Lab model card
First-Person Fairness evaluation 0.88% #5 of 10 Lab model card
World-Class Bio 46.4% #38 of 74 Official leaderboard
Molecular Biology Capabilities Test 48.1% #49 of 90 Official leaderboard
Agent JSK 99.8% #4 of 7 Lab model card
PatchEval 78.7% #4 of 7 Official leaderboard
PatchEval-Verified 78.7% #4 of 7 Official leaderboard
CyScenarioBench 11% #10 of 17 Lab model card
Human Pathogen Capabilities Test 47.3% #54 of 86 Official leaderboard
Multimodal Troubleshooting Virology 50.4% #17 of 27 Lab model card
BioTIER Permit 99.1% #46 of 69 Official leaderboard
ExploitGym 7.02% #12 of 18 Official leaderboard
ABC Bench Liquid-Handling Robot 72.6% #23 of 34 Official leaderboard
BinaryAudit 17.5% #14 of 20 Official leaderboard
DTap Prompt Injection 38.6% #6 of 8 Official leaderboard
LifeSciBench 47.9% #3 of 4 Official leaderboard
BiomniBench 55.2 #7 of 8 Official leaderboard
DTap Direct Misuse 48.9% #7 of 8 Official leaderboard
CoT-Control 0.30% #8 of 9 Lab model card
Tacit Knowledge and Troubleshooting 65% #27 of 30 Lab model card
PCT (Polarized Contrastive Pairs) 59.4% #42 of 46 Official leaderboard
PlugInject 97.8% #2 of 2 Lab model card

Other

BenchmarkScoreRankSource
ACCoRD 80.0 #13 of 152 Official leaderboard
Conceptual Reasoning Index 65.3 #22 of 151 Official leaderboard
LMCA 52% #26 of 147 Official leaderboard
CAIS Text Capabilities Index 41.3 #18 of 68 Official leaderboard
CAIS Vision Capabilities Index 57.7 #19 of 53 Official leaderboard
CAIS Risk Index 49.2 #30 of 67 Official leaderboard
SAGE (Vals) 43.3% #49 of 81 Official leaderboard

Its lab's numbers put it at 158.2, in line with its ECI

The Self-Reported Index fits one score to the 22 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 156.2 to 160.2. How we compute it.

Its lab's numbers next to 13 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
OpenAI MRCR v2 79.3% OpenAI MRCR v2 8-needle 128K-256K (36.6% at 512K-1M) 52.7% Leaderboard, 8-needle, 128K to 256K bin, medium reasoning +27 pts
Humanity's Last Exam 52.1% Humanity's Last Exam with tools 52.1% (without tools 39.8%) 36.2% Leaderboard, xhigh thinking +16 pts
MCP Atlas 67.2% MCP Atlas 70.6% Leaderboard, xhigh −3.4 pts
SWE-Bench Pro 57.7% SWE-Bench Pro (Public) 59.1% Leaderboard, xhigh, mini-SWE-agent −1.4 pts
Vals Finance Agent 56% FinanceAgent v1.1 (GPT-5.4) 57.1% Leaderboard, xhigh −1.1 pts
MMMU-Pro 81.2% MMMU Pro (no tools) 82.1% Leaderboard, thinking, with tools −0.9 pts
ARC-AGI-2 73.3% ARC-AGI-2 (Verified) 74% Leaderboard, xhigh −0.7 pts
GPQA Diamond 92.8% GPQA Diamond (GPT-5.4) 93.3% Epoch AI run, xhigh −0.5 pts
Online-Mind2Web 92.8% Online-Mind2Web, screenshot-based observations only 92.7% Leaderboard, ACT-2 Same
ARC-AGI 93.7% ARC-AGI-1 (Verified) 93.7% Leaderboard, xhigh Same
FrontierMath 47.6% FrontierMath Tier 1-3 (GPT-5.4) 47.6% Leaderboard, xhigh; task version 1.1.5, Epoch Same
FrontierMath Tier 4 27.1% FrontierMath Tier 4 (GPT-5.4) 27.1% Epoch AI run, xhigh Same
Toolathlon 54.6% Toolathlon 54.6% Leaderboard, xhigh Same

OpenAI's models over time

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

50 100 150 2023 2024 2025 2026 2027 GPT-6.1 Terra GPT-6 Luna GPT-5.4 Mini GPT-5.2 Pro GPT-5 GPT-5 nano GPT-4.1 nano 50 100 150 2023 2024 2025 2026 2027 GPT-6.1 Terra GPT-6 Luna

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-5.4 Mar 5, 2026

Its system card

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

GPT-5.4 Thinking System Card Mar 5, 2026

Health

Safety

Robustness

Bias

Alignment

Monitorability

Bio and chem risk

Cyber risk

AI R&D risk