DeepSeek

DeepSeek V3.2 Exp (Reasoning)

Unknown Size

By DeepSeek • Released 2025-09-29

Capability Radar

Avg Score
53

Across all benchmarks

Participated
12
Benchmarks

Benchmark Performance

Benchmark Category Score
AIME 2025
Reasoning
87.7
MMLU-Pro
Knowledge
85
GPQA Diamond
Knowledge
79.7
LiveCodeBench
Coding
78.9
LCR
Long-Context Reasoning
69
IFBench
Agent
54.1
SciCode
Reasoning Knowledge
37.7
𝜏²-Bench Telecom
Reasoning Knowledge
33.9
Artificial Analysis Coding Index
Coding
33.3
Artificial Analysis Intelligence Index
Knowledge
32.9
Terminal-Bench Hard
Agent Coding
31.1
HLE
Knowledge Multi-Modal
13.8