benchmark
.
space
benchmarks
rankings
compare
voices
transcripts
papers
articles
head to head
Gemini 3.1 Pro
vs
DeepSeek V4 Pro
9 shared benchmarks
7
wins
1
ties
1
wins
85.9%
BrowseComp
83.4%
94.3%
GPQA Diamond
90.1%
44.4%
Humanity's Last Exam
37.7%
78.2%
MCP-Atlas
73.6%
90.99%
MMLU Pro
87.5%
84.9%
MRCR v2
83.5%
54.2%
SWE-bench Pro
55.4%
80.6%
SWE-bench Verified
80.6%
68.5%
Terminal-Bench 2.0
67.9%