DeepSeek-V3.2
DeepSeek
Parameters
—
Architecture
—
Released
—
License
—
About
No detailed description available for this model.
Benchmark Scores
| Benchmark | Score | Date | Standing |
|---|---|---|---|
|
SWE-bench Verified
coding_agent
|
80.05%
|
03.02.2026 |
#39/60
better than 36%
|
|
SWE-bench Multilingual
coding_agent
|
67.81%
|
03.02.2026 |
#36/48
better than 26%
|
|
SWE-bench Pro
coding_agent
|
45.60%
|
03.02.2026 |
#58/77
better than 25%
|
|
Terminal-Bench 2.0
coding_agent
|
20.22%
|
03.02.2026 |
#18/23
better than 23%
|
|
Aider
coding_agent
|
100.00%
|
03.02.2026 |
#1/4
better than 100%
|
|
MMLU-Redux
knowledge
|
65.97%
|
26.06.2026 |
#22/30
better than 28%
|
|
MMLU-Pro
knowledge
|
17.74%
|
26.06.2026 |
#43/48
better than 11%
|
|
MMMLU
multilingual
|
89.52%
|
26.06.2026 |
#5/20
better than 79%
|
|
C-Eval
knowledge
|
75.23%
|
26.06.2026 |
#9/15
better than 43%
|
|
SuperGPQA
knowledge
|
42.34%
|
26.06.2026 |
#20/28
better than 30%
|
|
LongBench v2
long_context
|
34.62%
|
26.06.2026 |
#16/22
better than 29%
|
|
MMLU
knowledge
|
83.68%
|
26.06.2026 |
#13/17
better than 25%
|
What is a good SWE-bench Verified score? What is a good SWE-bench Multilingual score? What is a good SWE-bench Pro score?
No comments yet
Be the first to share your thoughts!