Anthropic Claude Haiku 4.5 vs Mistral Mistral 8x22B

Detailed comparison for LLMs

AnthropicMistral

On Guardion's LLM vulnerability Benchmark, Anthropic Claude Haiku 4.5 is the more secure of the two: Claude Haiku 4.5 scores 5.2% and Mistral 8x22B scores 31.5% 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 Haiku 4.5 is the overall winner in this comparison!

Attack Success Rate (lower is safer)

ASR for Anthropic Claude Haiku 4.5 vs Mistral Mistral 8x22B. Green marks the safer model on each metric. Only the overall score is available for estimated models.

Overall (ASR)

Claude Haiku 4.5
5.2%
Mistral 8x22B
31.5%

TAP Attack Method (ASR)

Claude Haiku 4.5
100.0%
Mistral 8x22B

Crescendo Attack Method (ASR)

Claude Haiku 4.5
100.0%
Mistral 8x22B

Zero-Shot (ASR)

Claude Haiku 4.5
100.0%
Mistral 8x22B
12.2%

Key Highlights

  • Anthropic Claude Haiku 4.5 has a lower Overall (ASR).
  • Mistral Mistral 8x22B has a lower Zero-Shot (ASR).

Security Profile

Outward is better on every axis.

OverallTAPCrescendoZero-Shot
Claude Haiku 4.5
Mistral 8x22B
Full security profile
Anthropic Claude Haiku 4.5
Full security profile
Mistral Mistral 8x22B

Frequently asked questions

Is Anthropic Claude Haiku 4.5 or Mistral Mistral 8x22B more secure?

On Guardion's LLM vulnerability Benchmark, Anthropic Claude Haiku 4.5 is the more secure of the two: Claude Haiku 4.5 scores 5.2% and Mistral 8x22B scores 31.5% 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 Haiku 4.5 vs Mistral 8x22B?

Claude Haiku 4.5 has a 5.2% ASR and Mistral 8x22B has a 31.5% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.

How were Claude Haiku 4.5 and Mistral 8x22B 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.

Related Comparisons