benchgap
OpenAI · model

GPT-6 Luna benchmark scores

As of 2026-11-22, GPT-6 Luna (OpenAI) has measured scores on 55 benchmarks and estimated scores on 120 more.

BenchmarkScoreSource
AA-GPQA Diamond92.8%estimated ± 1.1 pp, medium confidence
AA-HLE38.5%measured
AA-SciCode54.6%measured
AIME2697.9%estimated ± 1.4 pp, medium confidence
Artificial Analysis Intelligence Index27.9%measured
CritPt19.4%measured
GDPval-AA46.6%measured
GPQA Diamond94.8%estimated ± 1.2 pp, medium confidence
HLE54.2%estimated ± 2.2 pp, medium confidence
AA-LCR83.3%measured
LiveCodeBench v695.1%estimated ± 4.9 pp, medium confidence
AA-Omniscience Accuracy39.6%measured
SciCode54.3%estimated ± 3.1 pp, high confidence
Terminal-Bench 2.184.2%estimated ± 5.4 pp, low confidence
BrowseComp84.8%estimated ± 3.0 pp, medium confidence
HLE w/ tools56.9%estimated ± 3.1 pp, high confidence
LongBench v268.6%estimated ± 5.5 pp, medium confidence
AA Coding Index76.6%estimated ± 2.3 pp, low confidence
AA-MMMU-Pro79.7%measured
APEX-Agents39.0%estimated ± 9.4 pp, low confidence
APEX-Agents-AA37.9%estimated ± 9.3 pp, medium confidence
BioMysteryBench (human-difficult)47.9%estimated ± 6.8 pp, low confidence
DeepSearchQA89.3%estimated ± 9.6 pp, medium confidence
SWE-bench Verified86.5%estimated ± 5.1 pp, medium confidence
AutomationBench47.1%estimated ± 6.1 pp, low confidence
BFCL v476.4%estimated ± 2.7 pp, medium confidence
CyberGym79.4%estimated ± 6.0 pp, medium confidence
JobBench49.1%estimated ± 8.7 pp, medium confidence
SWE-bench Pro66.3%estimated ± 6.3 pp, medium confidence
WideResearch83.0%estimated ± 3.3 pp, medium confidence
AA Agentic Index40.7%estimated ± 3.1 pp, high confidence
MMLU-Pro85.5%estimated ± 5.0 pp, medium confidence
FailureSensorIQ66.5%estimated ± 1.7 pp, high confidence
GPQA93.6%estimated ± 0.3 pp, medium confidence
MMLU90.5%estimated ± 0.9 pp, medium confidence
Vals GPQA Diamond96.3%estimated ± 1.2 pp, low confidence
Vals LiveCodeBench86.0%estimated ± 5.7 pp, low confidence
Vals MMLU-Pro86.6%estimated ± 1.6 pp, medium confidence
MMMLU90.7%estimated ± 0.7 pp, low confidence
MMMU86.3%estimated ± 2.7 pp, high confidence
Chess Puzzles36.0%measured
AA MMLU-Pro87.1%estimated ± 0.8 pp, high confidence
Gert Labs58.8%measured
AA-AnalystAgent43.8%measured
AA EnterpriseOps-Gym47.6%estimated ± 7.2 pp, low confidence
AA Tau3 Banking56.2%estimated ± 2.4 pp, low confidence
AA Terminal-Bench 2.183.6%estimated ± 2.5 pp, high confidence
ApprenticeBench14.0%estimated ± 10.4 pp, low confidence
ARC-AGI-186.7%measured
ARC-AGI-259.3%measured
BabaIsBench86.2%estimated ± 14.1 pp, low confidence
CursorBench 3.155.3%estimated ± 3.4 pp, medium confidence
CursorBench 3.248.0%estimated ± 3.1 pp, low confidence
EBR-bench44.3%estimated ± 8.6 pp, low confidence
Terminal-Bench 3.017.1%estimated ± 3.2 pp, high confidence
FrontierCode 1.1 Main48.0%estimated ± 3.5 pp, low confidence
FrontierMath v2 (Tier 4)56.1%measured
FrontierMath v2 (Tiers 1-3)78.9%measured
FrontierSWE v236.5%estimated ± 8.0 pp, medium confidence
Furniture Assembly44.2%measured
MaxxBench51.7%measured
MazeBench0.0%measured
MirrorCode26.7%measured
Mystery Game Puzzles7.0%measured
NYT Connections68.7%measured
OfficeQA Pro54.1%estimated ± 6.8 pp, medium confidence
OSWorld-Verified80.1%estimated ± 8.4 pp, low confidence
ParseBench-164.1%measured
Terminal-Bench 4.115.2%measured
Terminal-Bench Hard56.7%estimated ± 2.8 pp, medium confidence
Vals SWE-bench85.0%estimated ± 8.2 pp, medium confidence
Terminal-Bench 2.1 (Vals AI)73.0%measured
VulcanBench v378.2%estimated ± 3.9 pp, medium confidence
CWE-Bench-AA56.7%measured
AA Cyber Index52.7%measured
DeepsecBench-AA23.5%measured
Harvey LAB-AA v1.14.3%estimated ± 3.5 pp, low confidence
HealthBench Professional (length-adjusted, Artificial Analysis)54.3%estimated ± 4.3 pp, low confidence
HealthBench Professional (raw, Artificial Analysis)65.1%estimated ± 0.8 pp, medium confidence
AA ITBench46.6%estimated ± 4.6 pp, medium confidence
Bug Hunt Bench18.3%measured
Chartography (tools)83.2%estimated ± 4.1 pp, low confidence
CursorBench 4.047.0%estimated ± 3.3 pp, high confidence
CWE-bench v151.7%estimated ± 4.5 pp, low confidence
DeepSWE66.6%measured
DRACO86.9%estimated ± 0.7 pp, low confidence
FrontierCode 2.022.7%measured
FrontierPhysics9.7%estimated ± 4.3 pp, low confidence
HLE w/o tools38.2%estimated ± 4.9 pp, high confidence
OSWorld 2.056.6%estimated ± 14.7 pp, low confidence
OSWorld 2.172.4%estimated ± 3.6 pp, high confidence
OSWorld 2.1 Offline48.9%measured
PostTrainBench v1.134.6%estimated ± 4.5 pp, high confidence
SpireClimb0.0%measured
StealthMark39.1%measured
SWE-bench Multimodal49.1%estimated ± 5.6 pp, low confidence
Terminal-Bench 4.019.2%estimated ± 3.7 pp, medium confidence
Terminal-Bench-Science 0.10.0%estimated ± 10.7 pp, low confidence
Toolathlon-Verified77.4%estimated ± 1.8 pp, medium confidence
VulcanBench Frontier v471.9%measured
AA AutomationBench53.2%measured
GDP.pdf22.8%measured
AA Terminal-Bench 4.012.6%measured
ARC-AGI-30.1%measured
ARC-AGI-3 Provider Adapter0.6%measured
cursorBench4144.8%estimated ± 3.7 pp, medium confidence
CWE-bench43.1%estimated ± 4.3 pp, low confidence
MKA36.8%measured
SkillsBench official leaderboard (BenchFlow)53.0%estimated ± 6.1 pp, medium confidence
BinaryAudit35.8%estimated ± 12.1 pp, low confidence
CompileBench55.6%estimated ± 10.0 pp, low confidence
OTelBench22.5%estimated ± 6.6 pp, medium confidence
FACTS Multimodal38.7%measured
Hermes Bench70.5%measured
Hermes Index33.9%measured
ParseBench-274.6%measured
CharXiv85.2%estimated ± 5.6 pp, low confidence
CharXiv w/o tools81.0%estimated ± 1.8 pp, medium confidence
ExploitBench33.7%estimated ± 1.9 pp, medium confidence
ExploitGym11.6%measured
AI2D_TEST94.1%estimated ± 2.0 pp, medium confidence
C-Eval93.9%estimated ± 1.0 pp, low confidence
CountBench98.0%estimated ± 2.9 pp, medium confidence
DynaMath85.5%estimated ± 3.0 pp, medium confidence
ERQA68.6%estimated ± 3.1 pp, high confidence
HMMT Feb 2026100.0%estimated ± 1.5 pp, low confidence
HMMT Nov 202595.8%estimated ± 1.5 pp, low confidence
MathVision90.3%estimated ± 3.1 pp, high confidence
MCP Atlas81.0%estimated ± 10.7 pp, low confidence
MMAnswerBench93.9%estimated ± 2.4 pp, low confidence
MMLU-Redux95.9%estimated ± 0.8 pp, medium confidence
MMMU-Pro80.3%estimated ± 1.5 pp, high confidence
NL2Repo59.4%estimated ± 6.0 pp, medium confidence
QwenClawBench93.2%estimated ± 6.0 pp, low confidence
RealWorldQA86.0%estimated ± 2.9 pp, medium confidence
ScreenSpot Pro85.1%estimated ± 10.1 pp, low confidence
SuperGPQA80.5%estimated ± 9.3 pp, medium confidence
Terminal-Bench 2.074.9%estimated ± 5.8 pp, medium confidence
Toolathlon55.9%estimated ± 6.3 pp, medium confidence
V*100.0%estimated ± 7.5 pp, medium confidence
Video-MME (w/o subtitle)86.5%estimated ± 1.7 pp, medium confidence
Video-MME (with subtitle)88.1%estimated ± 3.7 pp, medium confidence
HealthBench Hard31.4%measured
MedXpertQA (MM)76.9%estimated ± 6.5 pp, low confidence
React Native Evals78.6%estimated ± 4.7 pp, medium confidence
ResearchClawBench20.0%estimated ± 2.2 pp, medium confidence
PostTrainBench15.3%estimated ± 3.3 pp, low confidence
DAYJOB1.2%measured
DAYJOB Healthcare1.2%measured
Finance Agent v256.9%estimated ± 2.4 pp, medium confidence
BioMysteryBench (human-solvable)89.0%estimated ± 0.6 pp, medium confidence
HealthBench (raw)50.0%measured
HealthBench (length-adjusted)54.5%measured
HealthBench Professional60.8%measured
HealthBench Professional (raw)61.2%measured
HLE-Verified0.9%estimated ± 6.8 pp, low confidence
LABBench279.2%estimated ± 2.4 pp, low confidence
ScarfBench65.1%estimated ± 3.8 pp, medium confidence
BlenderWorld (text-only)70.0%estimated ± 2.8 pp, low confidence
Cookie-Bench76.9%estimated ± 1.3 pp, low confidence
STEMBench2.5%estimated ± 0.2 pp, medium confidence
SimpleQA45.0%estimated ± 4.3 pp, low confidence
Agents' Last Exam30.7%estimated ± 9.6 pp, medium confidence
BabyVision w/ Python90.7%estimated ± 4.0 pp, low confidence
OpenHarmony Bench v1.058.9%estimated ± 2.4 pp, high confidence
OpenHarmony Bench v1.255.5%estimated ± 11.0 pp, low confidence
SEC-Bench Pro34.2%measured
MRCRv285.9%estimated ± 11.3 pp, low confidence
MMLongBench-Doc61.5%estimated ± 2.6 pp, low confidence
SWE-Marathon41.5%estimated ± 2.1 pp, low confidence
BioBasicsBench37.6%measured
LifeVal76.2%estimated ± 2.3 pp, low confidence
MMMU-Pro w/ Python82.5%estimated ± 2.0 pp, high confidence
ExploitBench Internal Port0.0%measured
SWE-Atlas Codebase QnA59.5%estimated ± 4.8 pp, high confidence