Mistral Ministral 3 3B Reasoning 2512 vs DeepSeek DeepSeek-V3.2-Exp

Detailed comparison for LLMs

MistralDeepSeek

Mistral Ministral 3 3B Reasoning 2512 and DeepSeek DeepSeek-V3.2-Exp are evenly matched at about 46.0% attack success rate (ASR) on Guardion's LLM vulnerability Benchmark. One or both scores are estimated from public safety evaluations pending a Guardion benchmark run.

Head-to-Head Overview

It's a close match between Ministral 3 3B Reasoning 2512 and DeepSeek-V3.2-Exp!

Attack Success Rate (lower is safer)

ASR for Mistral Ministral 3 3B Reasoning 2512 vs DeepSeek DeepSeek-V3.2-Exp. Green marks the safer model on each metric. Only the overall score is available for estimated models.

Overall (ASR)

Ministral 3 3B Reasoning 2512
46.0%
DeepSeek-V3.2-Exp
46.0%

TAP Attack Method (ASR)

Ministral 3 3B Reasoning 2512
100.0%
DeepSeek-V3.2-Exp
100.0%

Crescendo Attack Method (ASR)

Ministral 3 3B Reasoning 2512
100.0%
DeepSeek-V3.2-Exp
100.0%

Zero-Shot (ASR)

Ministral 3 3B Reasoning 2512
100.0%
DeepSeek-V3.2-Exp
100.0%

Key Highlights

No specific highlights generated for this comparison.

Security Profile

Outward is better on every axis.

OverallTAPCrescendoZero-Shot
Ministral 3 3B Reasoning 2512
DeepSeek-V3.2-Exp
Full security profile
Mistral Ministral 3 3B Reasoning 2512
Full security profile
DeepSeek DeepSeek-V3.2-Exp

Frequently asked questions

Is Mistral Ministral 3 3B Reasoning 2512 or DeepSeek DeepSeek-V3.2-Exp more secure?

Mistral Ministral 3 3B Reasoning 2512 and DeepSeek DeepSeek-V3.2-Exp are evenly matched at about 46.0% attack success rate (ASR) on Guardion's LLM vulnerability Benchmark. One or both scores are estimated from public safety evaluations pending a Guardion benchmark run.

What is the attack success rate (ASR) of Ministral 3 3B Reasoning 2512 vs DeepSeek-V3.2-Exp?

Ministral 3 3B Reasoning 2512 has a 46.0% ASR and DeepSeek-V3.2-Exp has a 46.0% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.

How were Ministral 3 3B Reasoning 2512 and DeepSeek-V3.2-Exp 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