DeepSeek V4 Pro
DeepSeek
Parameters
—
Architecture
—
Released
—
License
—
About
No detailed description available for this model.
Benchmark Scores
| Benchmark | Score | Date | Standing |
|---|---|---|---|
|
BioMysteryBench
stem_reasoning
|
26.07%
|
28.08.2026 |
#7/10
better than 33%
|
|
SWE-bench Verified
coding_agent
|
91.97%
|
11.06.2026 |
#7/60
better than 86%
|
|
Terminal Bench 2.1
coding_agent
|
65.63%
|
11.06.2026 |
#28/44
better than 37%
|
|
VIBE-V2
coding_agent
|
60.93%
|
11.06.2026 |
#6/9
better than 38%
|
|
SVG-Bench
coding_agent
|
85.09%
|
11.06.2026 |
#4/9
better than 62%
|
|
GDPVal
general_agent
|
71.12%
|
11.06.2026 |
#6/20
better than 74%
|
|
SpreadSheetBench-v1
general_agent
|
91.76%
|
11.06.2026 |
#5/9
better than 38%
|
|
YC-Bench
general_agent
|
22.73%
|
11.06.2026 |
#6/8
better than 29%
|
|
MCP-Atlas
general_agent
|
79.63%
|
11.06.2026 |
#22/54
better than 58%
|
|
Apex-Agents
general_agent
|
41.74%
|
28.08.2026 |
#17/24
better than 30%
|
|
HLE (with tools)
stem_reasoning
|
82.17%
|
28.08.2026 |
#7/49
better than 85%
|
|
SWE-Marathon
coding_agent
|
35.50%
|
28.08.2026 |
#12/17
better than 31%
|
|
BrowseComp
general_agent
|
90.22%
|
11.06.2026 |
#9/45
better than 82%
|
|
SWE-bench Multilingual
coding_agent
|
85.56%
|
28.08.2026 |
#11/48
better than 79%
|
|
SWE Atlas - QnA
coding_agent
|
80.66%
|
28.08.2026 |
#7/24
better than 74%
|
|
SWE Atlas - TW
coding_agent
|
64.07%
|
28.08.2026 |
#6/20
better than 74%
|
|
SWE Atlas - RF
coding_agent
|
79.53%
|
28.08.2026 |
#6/14
better than 62%
|
|
Terminal-Bench 2.1 (Terminus-2)
coding_agent
|
98.67%
|
28.08.2026 |
#4/44
better than 93%
|
|
NL2Repo
coding_agent
|
78.77%
|
28.08.2026 |
#4/41
better than 90%
|
|
Cybergym
general_agent
|
90.28%
|
28.08.2026 |
#4/17
better than 75%
|
|
ProgramBench
coding_agent
|
18.14%
|
28.08.2026 |
#15/18
better than 12%
|
|
PostTrainBench
coding_agent
|
38.91%
|
28.08.2026 |
#18/22
better than 19%
|
|
Harbor-Index
coding_agent
|
51.57%
|
28.08.2026 |
#6/7
better than 17%
|
|
Hy-Backend 2.0 (Internal)
coding_agent
|
32.48%
|
28.08.2026 |
#7/8
better than 14%
|
|
Hy-CompanyBench V2 (Internal)
general_agent
|
80.65%
|
28.08.2026 |
#4/8
better than 57%
|
|
WideSearch
general_agent
|
91.19%
|
28.08.2026 |
#7/24
better than 74%
|
|
$OneMillion-Bench
general_capabilities
|
76.81%
|
28.08.2026 |
#5/11
better than 60%
|
|
DRACO
general_agent
|
51.71%
|
28.08.2026 |
#6/12
better than 55%
|
|
Hy-LifeSearch (Internal)
general_agent
|
32.65%
|
28.08.2026 |
#6/8
better than 29%
|
|
Hy-BrowseComp-Pro2 (Internal)
general_agent
|
46.50
|
28.08.2026 |
#8/8
better than 0%
|
|
OfficeQA Pro
general_agent
|
95.94%
|
28.08.2026 |
#5/14
better than 54%
|
|
Toolathlon Verified
general_agent
|
92.42%
|
28.08.2026 |
#8/35
better than 74%
|
|
SkillsBench Avg5
coding_agent
|
97.27%
|
28.08.2026 |
#2/16
better than 93%
|
|
JobBench
general_agent
|
69.91%
|
28.08.2026 |
#8/19
better than 61%
|
|
WorkSpaceBench
general_agent
|
42.86%
|
28.08.2026 |
#7/11
better than 40%
|
|
GDPVal-AA v2
general_agent
|
86.29%
|
28.08.2026 |
#15/38
better than 62%
|
|
BankerToolBench
general_agent
|
51.11%
|
28.08.2026 |
#11/15
better than 29%
|
|
E-Bench (Internal)
general_agent
|
39.88%
|
28.08.2026 |
#7/8
better than 14%
|
|
E-Bench-Code (Internal)
coding_agent
|
64.30
|
28.08.2026 |
#8/8
better than 0%
|
|
Hy-FinAgentBench (Internal)
domain_finance
|
66.67%
|
28.08.2026 |
#5/8
better than 29%
|
|
Hy-FinmodelBench v2 (Internal)
domain_finance
|
60.70%
|
28.08.2026 |
#7/8
better than 14%
|
|
CritPt (no tools)
stem_reasoning
|
54.89%
|
28.08.2026 |
#9/21
better than 60%
|
|
GPQA Diamond
stem_reasoning
|
97.16%
|
28.08.2026 |
#8/80
better than 90%
|
|
Humanity's Last Exam
stem_reasoning
|
76.89%
|
28.08.2026 |
#12/71
better than 83%
|
|
SUPERChem
stem_reasoning
|
39.00%
|
28.08.2026 |
#5/8
better than 43%
|
|
ArXivMath
stem_reasoning
|
37.41%
|
28.08.2026 |
#5/7
better than 33%
|
|
HorizonMath (pass@4)
stem_reasoning
|
12.92%
|
28.08.2026 |
#6/7
better than 17%
|
|
MathArena Apex 2025
stem_reasoning
|
52.37%
|
28.08.2026 |
#6/8
better than 29%
|
|
BrokenArXiv
stem_reasoning
|
32.16%
|
28.08.2026 |
#5/7
better than 33%
|
|
Claw-Eval Avg
coding_agent
|
69.89%
|
11.06.2026 |
#24/40
better than 41%
|
|
DeepSWE
coding_agent
|
86.24%
|
28.08.2026 |
#8/21
better than 60%
|
|
Hy-SWE Max Verified (Internal)
coding_agent
|
79.15%
|
28.08.2026 |
#4/8
better than 57%
|
|
Agents' Last Exam
general_agent
|
34.66%
|
28.08.2026 |
#22/29
better than 25%
|
|
NL2Repo-Bench
coding_agent
|
25.88%
|
11.06.2026 |
#21/23
better than 9%
|
|
Automation-Bench
general_agent
|
37.69%
|
28.08.2026 |
#14/27
better than 50%
|
|
SWE-bench Pro
coding_agent
|
61.76%
|
11.06.2026 |
#33/77
better than 55%
|
What is a good BioMysteryBench score? What is a good SWE-bench Verified score? What is a good Terminal Bench 2.1 score?
No comments yet
Be the first to share your thoughts!