Claude Opus 4.7 — Real Scores Across 17 Benchmarks (2026)

Anthropic

Claude Opus 4.7

17 benchmarks

HumanEval

#2 of 18

96.3%

GPQA Diamond

#3 of 95

94.2%

MMMLU

#1 of 6

92.0%

BigLaw Bench

#2 of 3

90.9%

CharXiv

#1 of 10

89.0%

SWE-bench Verified

#2 of 77

87.6%

MMLU Pro

#9 of 50

87.0%

SWE-bench Multilingual

#2 of 12

85.7%

LiveCodeBench

#13 of 49

83.0%

BrowseComp

#9 of 28

79.3%

Aider Polyglot

#7 of 12

79.0%

OSWorld

#3 of 30

78.0%

Terminal-Bench 2.0

#7 of 24

77.0%

CyberGym

#2 of 4

73.8%

CursorBench

#1 of 1

70.0%

SWE-bench Pro

#2 of 22

64.3%

SWE-bench Multimodal

#2 of 3

35.0%