benchmark
.
space
benchmarks
rankings
compare
voices
transcripts
papers
articles
head to head
Kimi K2.5
vs
DeepSeek V4 Pro
9 shared benchmarks
0
wins
0
ties
9
wins
60.6%
BrowseComp
83.4%
87.6%
GPQA Diamond
90.1%
24.37%
Humanity's Last Exam
37.7%
85.0%
LiveCodeBench
93.5%
87.1%
MMLU Pro
87.5%
73.0%
SWE-bench Multilingual
76.2%
50.7%
SWE-bench Pro
55.4%
76.8%
SWE-bench Verified
80.6%
50.8%
Terminal-Bench 2.0
67.9%