DeepSeek

DeepSeek V3.2 (Reasoning)

Unknown Size

By DeepSeek • Released 2025-12-01

Capability Radar

Avg Score
62

Across all benchmarks

Participated
13
Benchmarks

Benchmark Performance

Benchmark Category Score
AIME 2025
Reasoning
92
𝜏²-Bench Telecom
Reasoning Knowledge
90.6
LiveCodeBench
Coding
86.2
MMLU-Pro
Knowledge
86.2
GPQA Diamond
Knowledge
84
LCR
Long-Context Reasoning
65
IFBench
Agent
60.7
SWE-bench (Bash Only)
Coding Agent
60
Artificial Analysis Intelligence Index
Knowledge
41.6
SciCode
Reasoning Knowledge
38.9
Artificial Analysis Coding Index
Coding
36.7
Terminal-Bench Hard
Agent Coding
35.6
HLE
Knowledge Multi-Modal
22.2