DeepSeek V4 Pro
16 benchmarks
HMMT Feb 2026
#3 of 5
95.2%
LiveCodeBench
#2 of 53
93.5%
GPQA Diamond
#29 of 118
90.1%
MMLU Pro
#6 of 55
87.5%
MRCR v2
#4 of 18
83.5%
BrowseComp
#17 of 45
83.4%
SWE-bench Verified
#6 of 86
80.6%
SWE-bench Multilingual
#6 of 16
76.2%
MCP-Atlas
#15 of 17
73.6%
Terminal-Bench 2.0
#12 of 29
67.9%
SimpleQA Verified
#2 of 5
57.9%
SWE-bench Pro
#24 of 43
55.4%
Toolathlon
#5 of 7
51.8%
Humanity's Last Exam
#23 of 72
48.2%
APEX-Agents
#6 of 7
38.3%
Humanity's Last Exam
#23 of 72
37.7%