Moonshot AI Kimi K2-Thinking-0905 vs Meta Llama 3H 405B
Detailed comparison for LLMs
On Guardion's LLM vulnerability Benchmark, Meta Llama 3H 405B is the more secure of the two: Kimi K2-Thinking-0905 scores 38.0% and Llama 3H 405B scores 36.6% 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
Llama 3H 405B is the overall winner in this comparison!
Attack Success Rate (lower is safer)
ASR for Moonshot AI Kimi K2-Thinking-0905 vs Meta Llama 3H 405B. 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
- Meta Llama 3H 405B has a lower Overall (ASR).
- Meta Llama 3H 405B has a lower Zero-Shot (ASR).
Security Profile
Outward is better on every axis.
Frequently asked questions
Is Moonshot AI Kimi K2-Thinking-0905 or Meta Llama 3H 405B more secure?
On Guardion's LLM vulnerability Benchmark, Meta Llama 3H 405B is the more secure of the two: Kimi K2-Thinking-0905 scores 38.0% and Llama 3H 405B scores 36.6% 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 Kimi K2-Thinking-0905 vs Llama 3H 405B?
Kimi K2-Thinking-0905 has a 38.0% ASR and Llama 3H 405B has a 36.6% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.
How were Kimi K2-Thinking-0905 and Llama 3H 405B 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.