DeepSeek V4 Flash
13 benchmarks
AIME
#23 of 117
94.8%
LiveCodeBench
#4 of 53
91.6%
GPQA Diamond
#36 of 118
88.1%
MMLU Pro
#11 of 55
86.2%
SWE-bench Verified
#14 of 86
79.0%
MRCR v2
#5 of 18
78.7%
SWE-bench Multilingual
#8 of 16
73.3%
BrowseComp
#31 of 45
73.2%
Terminal-Bench 2.1
#23 of 27
62.0%
Terminal-Bench 2.0
#21 of 29
56.9%
SWE-bench Pro
#32 of 43
52.6%
Humanity's Last Exam
#46 of 72
34.8%
SimpleQA Verified
#4 of 5
34.1%