benchgap
Anthropic · model

Claude 3 Opus benchmark scores

As of 2026-11-22, Claude 3 Opus (Anthropic) has measured scores on 5 benchmarks and estimated scores on 56 more.

BenchmarkScoreSource
AA-GPQA Diamond48.9%measured
AA-HLE2.8%measured
AA-SciCode31.1%estimated ± 3.4 pp, high confidence
Artificial Analysis Intelligence Index8.7%measured
CritPt0.1%estimated ± 3.4 pp, high confidence
GPQA Diamond67.6%estimated ± 2.4 pp, medium confidence
HLE16.4%estimated ± 8.5 pp, low confidence
LiveCodeBench v670.8%estimated ± 4.7 pp, high confidence
AA-Omniscience Accuracy19.4%estimated ± 7.3 pp, medium confidence
SciCode31.8%estimated ± 4.6 pp, high confidence
AA Coding Index19.5%measured
BioMysteryBench (human-difficult)32.9%estimated ± 8.3 pp, low confidence
SWE-bench Verified42.6%estimated ± 6.5 pp, low confidence
SWE-bench Pro26.4%estimated ± 6.0 pp, low confidence
MMLU-Pro76.6%estimated ± 5.4 pp, medium confidence
FailureSensorIQ63.6%estimated ± 2.1 pp, medium confidence
GPQA53.3%estimated ± 6.1 pp, medium confidence
MMLU83.7%estimated ± 3.8 pp, high confidence
Vals GPQA Diamond48.0%estimated ± 3.2 pp, high confidence
Vals LiveCodeBench53.1%estimated ± 7.1 pp, medium confidence
Vals MMLU-Pro68.1%estimated ± 3.0 pp, high confidence
MMMLU77.1%estimated ± 1.7 pp, medium confidence
Chess Puzzles4.0%measured
AA MMLU-Pro83.5%estimated ± 1.2 pp, medium confidence
ARC-AGI-172.8%estimated ± 7.9 pp, low confidence
ARC-AGI-226.1%estimated ± 14.4 pp, low confidence
EBR-bench0.5%estimated ± 12.7 pp, low confidence
Furniture Assembly24.3%estimated ± 14.1 pp, low confidence
MaxxBench30.3%estimated ± 10.3 pp, low confidence
Mystery Game Puzzles2.4%estimated ± 10.3 pp, low confidence
NYT Connections21.2%estimated ± 12.6 pp, low confidence
Vals SWE-bench43.3%estimated ± 5.5 pp, low confidence
HealthBench Professional (length-adjusted, Artificial Analysis)48.4%estimated ± 4.3 pp, low confidence
HealthBench Professional (raw, Artificial Analysis)38.8%estimated ± 1.1 pp, low confidence
FrontierPhysics0.0%estimated ± 6.1 pp, low confidence
HLE w/o tools22.8%estimated ± 4.9 pp, high confidence
SpireClimb0.0%estimated ± 10.3 pp, low confidence
StealthMark28.0%estimated ± 9.7 pp, low confidence
ARC-AGI-31.3%estimated ± 9.8 pp, low confidence
ARC-AGI-3 Provider Adapter2.7%estimated ± 7.3 pp, low confidence
SkillsBench official leaderboard (BenchFlow)23.7%estimated ± 6.2 pp, low confidence
Vibe Code Bench5.3%estimated ± 10.0 pp, low confidence
FACTS Multimodal30.3%estimated ± 5.6 pp, low confidence
ExploitBench3.7%estimated ± 13.4 pp, low confidence
C-Eval88.2%estimated ± 1.0 pp, low confidence
MMLU-Redux89.4%estimated ± 1.2 pp, low confidence
NL2Repo11.2%estimated ± 4.7 pp, medium confidence
SuperGPQA61.7%estimated ± 10.1 pp, low confidence
AA LiveCodeBench84.0%estimated ± 4.7 pp, low confidence
React Native Evals69.0%estimated ± 3.2 pp, medium confidence
DAYJOB0.4%estimated ± 6.3 pp, low confidence
DAYJOB Healthcare0.1%estimated ± 6.1 pp, low confidence
HealthBench (length-adjusted)54.1%estimated ± 3.7 pp, medium confidence
HealthBench Professional56.7%estimated ± 4.2 pp, low confidence
ScarfBench24.4%estimated ± 9.1 pp, low confidence
BlenderWorld (text-only)70.6%estimated ± 4.0 pp, low confidence
Cookie-Bench72.2%estimated ± 2.5 pp, low confidence
SimpleQA12.9%estimated ± 7.8 pp, medium confidence
SEC-Bench Pro6.9%estimated ± 12.3 pp, low confidence
BioBasicsBench11.1%estimated ± 10.2 pp, low confidence
ExploitBench Internal Port0.0%estimated ± 1.4 pp, low confidence