GPT-5.5 vs Claude Opus 4.7 — Who Actually Wins? Real Benchmark Data (2026)

head to head

GPT-5.5 vs Claude Opus 4.7

16 shared benchmarks

13

wins

0

ties

3

wins

85.0%

ARC-AGI 2

75.8%

84.4%

BrowseComp

79.3%

84.1%

CharXiv

82.1%

81.8%

CyberGym

73.8%

51.7%

FrontierMath

43.8%

93.5%

GPQA Diamond

94.2%

41.4%

Humanity's Last Exam

46.9%

82.8%

MCP-Atlas

79.1%

81.2%

MMMU Pro

75.2%

74.0%

MRCR v2

59.2%

78.7%

OSWorld

78.0%

49.5%

OSWorld 2.0

48.9%

78.7%

OSWorld-Verified

78.0%

58.6%

SWE-bench Pro

64.3%

82.7%

Terminal-Bench 2.0

77.0%

83.4%

Terminal-Bench 2.1

66.1%