DeepSeek

DeepSeek V3.2 Speciale

Unknown Size

By DeepSeek • Released 2025-12-01

Capability Radar

Avg Score
55

Across all benchmarks

Participated
12
Benchmarks

Benchmark Performance

Benchmark Category Score
AIME 2025
Reasoning
96.7
LiveCodeBench
Coding
89.6
GPQA Diamond
Knowledge
87.1
MMLU-Pro
Knowledge
86.3
IFBench
Agent
63.9
LCR
Long-Context Reasoning
59.3
SciCode
Reasoning Knowledge
44
Artificial Analysis Coding Index
Coding
37.9
Terminal-Bench Hard
Agent Coding
34.8
Artificial Analysis Intelligence Index
Knowledge
34.1
HLE
Knowledge Multi-Modal
26.1
𝜏²-Bench Telecom
Reasoning Knowledge
0