Anthropic Claude 3.7 Sonnet vs Microsoft Phi-4
Detailed comparison for LLMs
On Guardion's LLM vulnerability Benchmark, Anthropic Claude 3.7 Sonnet is the more secure of the two: Claude 3.7 Sonnet scores 20.3% and Phi-4 scores 30.0% on attack success rate (ASR) (lower is better). One or both scores are estimated from public safety evaluations pending a Guardion benchmark run.
Head-to-Head Overview
Claude 3.7 Sonnet is the overall winner in this comparison!
Attack Success Rate (lower is safer)
ASR for Anthropic Claude 3.7 Sonnet vs Microsoft Phi-4. Green marks the safer model on each metric. Only the overall score is available for estimated models.
Overall (ASR)
TAP Attack Method (ASR)
Crescendo Attack Method (ASR)
Key Highlights
- Anthropic Claude 3.7 Sonnet has a lower Overall (ASR).
Security Profile
Outward is better on every axis.
Frequently asked questions
Is Anthropic Claude 3.7 Sonnet or Microsoft Phi-4 more secure?
On Guardion's LLM vulnerability Benchmark, Anthropic Claude 3.7 Sonnet is the more secure of the two: Claude 3.7 Sonnet scores 20.3% and Phi-4 scores 30.0% on attack success rate (ASR) (lower is better). One or both scores are estimated from public safety evaluations pending a Guardion benchmark run.
What is the attack success rate (ASR) of Claude 3.7 Sonnet vs Phi-4?
Claude 3.7 Sonnet has a 20.3% ASR and Phi-4 has a 30.0% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.
How were Claude 3.7 Sonnet and Phi-4 tested?
Both were red-teamed with the HarmBench framework across zero-shot, TAP (Tree of Attacks with Pruning), and Crescendo multi-turn attacks, scored by Attack Success Rate.