GPT-5.5 vs Claude Opus 4.6 — Who Actually Wins? Real Benchmark Data (2026)

head to head

GPT-5.5 vs Claude Opus 4.6

11 shared benchmarks

8

wins

1

ties

2

wins

85.0%

ARC-AGI 2

85.0%

0.43%

ARC-AGI 3

0.5%

84.4%

BrowseComp

84.0%

84.1%

CharXiv

61.5%

81.8%

CyberGym

66.6%

93.5%

GPQA Diamond

91.3%

41.4%

Humanity's Last Exam

40.0%

74.0%

MRCR v2

93.0%

78.7%

OSWorld

72.7%

58.6%

SWE-bench Pro

53.4%

82.7%

Terminal-Bench 2.0

65.4%