Gemini 3.1 Pro vs GPT-5.4
18 shared benchmarks
9
wins
0
ties
9
wins
77.1%
ARC-AGI 2
73.3%
0.37%
ARC-AGI 3
0.26%
85.9%
BrowseComp
82.7%
12.0%
DeepSWE
52.0%
36.9%
FrontierMath
47.6%
94.3%
GPQA Diamond
92.8%
44.4%
Humanity's Last Exam
36.24%
80.5%
MMMU Pro
81.2%
84.9%
MRCR v2
36.6%
72.7%
OSWorld
75.0%
76.2%
OSWorld-Verified
75.0%
57.3%
Online-Mind2Web
92.8%
54.2%
SWE-bench Pro
57.7%
80.6%
SWE-bench Verified
77.2%
68.5%
Terminal-Bench 2.0
75.1%
74.4%
USAMO 2026
95.2%
53.8%
WebArena
67.3%
88.6%
WebVoyager
70.9%