GPT-5.5 vs Gemini 3.1 Pro — Who Actually Wins? Real Benchmark Data (2026)

head to head

GPT-5.5 vs Gemini 3.1 Pro

17 shared benchmarks

13

wins

0

ties

4

wins

85.0%

ARC-AGI 2

77.1%

0.43%

ARC-AGI 3

0.37%

84.4%

BrowseComp

85.9%

84.1%

CharXiv

83.3%

67.0%

DeepSWE

12.0%

51.7%

FrontierMath

36.9%

93.5%

GPQA Diamond

94.3%

41.4%

Humanity's Last Exam

44.4%

82.8%

MCP-Atlas

78.2%

81.2%

MMMU Pro

80.5%

74.0%

MRCR v2

84.9%

78.7%

OSWorld

72.7%

78.7%

OSWorld-Verified

76.2%

58.6%

SWE-bench Pro

54.2%

82.7%

Terminal-Bench 2.0

68.5%

83.4%

Terminal-Bench 2.1

70.3%

98.0%

τ2-bench

90.8%