Alibaba Qwen2 7B Instruct vs DeepSeek DeepSeek R1 Distill Llama 8B

Detailed comparison for LLMs

AlibabaDeepSeek

On Guardion's LLM vulnerability Benchmark, DeepSeek DeepSeek R1 Distill Llama 8B is the more secure of the two: Qwen2 7B Instruct scores 52.0% and DeepSeek R1 Distill Llama 8B scores 48.0% 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

DeepSeek R1 Distill Llama 8B is the overall winner in this comparison!

Attack Success Rate (lower is safer)

ASR for Alibaba Qwen2 7B Instruct vs DeepSeek DeepSeek R1 Distill Llama 8B. Green marks the safer model on each metric. Only the overall score is available for estimated models.

Overall (ASR)

Qwen2 7B Instruct
52.0%
DeepSeek R1 Distill Llama 8B
48.0%

TAP Attack Method (ASR)

Qwen2 7B Instruct
100.0%
DeepSeek R1 Distill Llama 8B
100.0%

Crescendo Attack Method (ASR)

Qwen2 7B Instruct
100.0%
DeepSeek R1 Distill Llama 8B
100.0%

Zero-Shot (ASR)

Qwen2 7B Instruct
100.0%
DeepSeek R1 Distill Llama 8B
100.0%

Key Highlights

  • DeepSeek DeepSeek R1 Distill Llama 8B has a lower Overall (ASR).

Security Profile

Outward is better on every axis.

OverallTAPCrescendoZero-Shot
Qwen2 7B Instruct
DeepSeek R1 Distill Llama 8B
Full security profile
Alibaba Qwen2 7B Instruct
Full security profile
DeepSeek DeepSeek R1 Distill Llama 8B

Frequently asked questions

Is Alibaba Qwen2 7B Instruct or DeepSeek DeepSeek R1 Distill Llama 8B more secure?

On Guardion's LLM vulnerability Benchmark, DeepSeek DeepSeek R1 Distill Llama 8B is the more secure of the two: Qwen2 7B Instruct scores 52.0% and DeepSeek R1 Distill Llama 8B scores 48.0% 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 Qwen2 7B Instruct vs DeepSeek R1 Distill Llama 8B?

Qwen2 7B Instruct has a 52.0% ASR and DeepSeek R1 Distill Llama 8B has a 48.0% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.

How were Qwen2 7B Instruct and DeepSeek R1 Distill Llama 8B 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