Anthropic Claude Sonnet 4.5 vs Anthropic Claude Haiku 4.5
Detailed comparison for LLMs
On Guardion's LLM vulnerability Benchmark, Anthropic Claude Sonnet 4.5 is the more secure of the two: Claude Sonnet 4.5 scores 4.2% and Claude Haiku 4.5 scores 5.2% 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 Sonnet 4.5 is the overall winner in this comparison!
Attack Success Rate (lower is safer)
ASR for Anthropic Claude Sonnet 4.5 vs Anthropic Claude Haiku 4.5. 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)
Zero-Shot (ASR)
Key Highlights
- Anthropic Claude Sonnet 4.5 has a lower Overall (ASR).
Security Profile
Outward is better on every axis.
Frequently asked questions
Is Anthropic Claude Sonnet 4.5 or Anthropic Claude Haiku 4.5 more secure?
On Guardion's LLM vulnerability Benchmark, Anthropic Claude Sonnet 4.5 is the more secure of the two: Claude Sonnet 4.5 scores 4.2% and Claude Haiku 4.5 scores 5.2% 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 Sonnet 4.5 vs Claude Haiku 4.5?
Claude Sonnet 4.5 has a 4.2% ASR and Claude Haiku 4.5 has a 5.2% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.
How were Claude Sonnet 4.5 and Claude Haiku 4.5 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.