Anthropic Claude 3.5 Sonnet v1 vs Meta Llama 3.1 405B Instruct

Detailed comparison for LLMs

AnthropicMeta

On Guardion's LLM vulnerability Benchmark, Anthropic Claude 3.5 Sonnet v1 is the more secure of the two: Claude 3.5 Sonnet v1 scores 16.0% and Llama 3.1 405B Instruct scores 36.0% 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 3.5 Sonnet v1 is the overall winner in this comparison!

Attack Success Rate (lower is safer)

ASR for Anthropic Claude 3.5 Sonnet v1 vs Meta Llama 3.1 405B Instruct. Green marks the safer model on each metric. Only the overall score is available for estimated models.

Overall (ASR)

Claude 3.5 Sonnet v1
16.0%
Llama 3.1 405B Instruct
36.0%

TAP Attack Method (ASR)

Claude 3.5 Sonnet v1
28.4%
Llama 3.1 405B Instruct
100.0%

Crescendo Attack Method (ASR)

Claude 3.5 Sonnet v1
19.3%
Llama 3.1 405B Instruct
100.0%

Zero-Shot (ASR)

Claude 3.5 Sonnet v1
0.3%
Llama 3.1 405B Instruct
100.0%

Key Highlights

  • Anthropic Claude 3.5 Sonnet v1 has a lower Overall (ASR).
  • Anthropic Claude 3.5 Sonnet v1 has a lower TAP Attack Method (ASR).
  • Anthropic Claude 3.5 Sonnet v1 has a lower Crescendo Attack Method (ASR).
  • Anthropic Claude 3.5 Sonnet v1 has a lower Zero-Shot (ASR).

Security Profile

Outward is better on every axis.

OverallTAPCrescendoZero-Shot
Claude 3.5 Sonnet v1
Llama 3.1 405B Instruct
Full security profile
Anthropic Claude 3.5 Sonnet v1
Full security profile
Meta Llama 3.1 405B Instruct

Frequently asked questions

Is Anthropic Claude 3.5 Sonnet v1 or Meta Llama 3.1 405B Instruct more secure?

On Guardion's LLM vulnerability Benchmark, Anthropic Claude 3.5 Sonnet v1 is the more secure of the two: Claude 3.5 Sonnet v1 scores 16.0% and Llama 3.1 405B Instruct scores 36.0% 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 3.5 Sonnet v1 vs Llama 3.1 405B Instruct?

Claude 3.5 Sonnet v1 has a 16.0% ASR and Llama 3.1 405B Instruct has a 36.0% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.

How were Claude 3.5 Sonnet v1 and Llama 3.1 405B Instruct 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