Anthropic Claude 4.0 Sonnet vs DeepSeek DeepSeek-V3.2 Non-thinking
Detailed comparison for LLMs
AnthropicDeepSeek
Head-to-Head Overview
DeepSeek-V3.2 Non-thinking is the overall winner in this comparison!
LLM Metric
Comparison of Attack Success Rate (ASR) metrics for Anthropic Claude 4.0 Sonnet and DeepSeek DeepSeek-V3.2 Non-thinking across the three attack methods.
Overall (ASR)
Claude 4.0 Sonnet
13.630
DeepSeek-V3.2 Non-thinking
0
TAP Attack Method (ASR)
Claude 4.0 Sonnet
22.020
DeepSeek-V3.2 Non-thinking
0
Crescendo Attack Method (ASR)
Claude 4.0 Sonnet
14.790
DeepSeek-V3.2 Non-thinking
0
Zero-Shot (ASR)
Claude 4.0 Sonnet
4.080
DeepSeek-V3.2 Non-thinking
0
Key Highlights
- DeepSeek DeepSeek-V3.2 Non-thinking has a lower Overall (ASR).
- DeepSeek DeepSeek-V3.2 Non-thinking has a lower TAP Attack Method (ASR).
- DeepSeek DeepSeek-V3.2 Non-thinking has a lower Crescendo Attack Method (ASR).
- DeepSeek DeepSeek-V3.2 Non-thinking has a lower Zero-Shot (ASR).
Metric Radar Chart
claude-4-0-sonnet
deepseek-v3-2-non-thinking