Google Gemma 3 1B vs Mistral Ministral 3 14B Base 2512

Detailed comparison for LLMs

GoogleMistral

On Guardion's LLM vulnerability Benchmark, Mistral Ministral 3 14B Base 2512 is the more secure of the two: Gemma 3 1B scores 60.0% and Ministral 3 14B Base 2512 scores 58.0% 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

Ministral 3 14B Base 2512 is the overall winner in this comparison!

Attack Success Rate (lower is safer)

ASR for Google Gemma 3 1B vs Mistral Ministral 3 14B Base 2512. Green marks the safer model on each metric. Only the overall score is available for estimated models.

Overall (ASR)

Gemma 3 1B
60.0%
Ministral 3 14B Base 2512
58.0%

TAP Attack Method (ASR)

Gemma 3 1B
100.0%
Ministral 3 14B Base 2512
100.0%

Crescendo Attack Method (ASR)

Gemma 3 1B
100.0%
Ministral 3 14B Base 2512
100.0%

Zero-Shot (ASR)

Gemma 3 1B
100.0%
Ministral 3 14B Base 2512
100.0%

Key Highlights

  • Mistral Ministral 3 14B Base 2512 has a lower Overall (ASR).

Security Profile

Outward is better on every axis.

OverallTAPCrescendoZero-Shot
Gemma 3 1B
Ministral 3 14B Base 2512
Full security profile
Google Gemma 3 1B
Full security profile
Mistral Ministral 3 14B Base 2512

Frequently asked questions

Is Google Gemma 3 1B or Mistral Ministral 3 14B Base 2512 more secure?

On Guardion's LLM vulnerability Benchmark, Mistral Ministral 3 14B Base 2512 is the more secure of the two: Gemma 3 1B scores 60.0% and Ministral 3 14B Base 2512 scores 58.0% 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 Gemma 3 1B vs Ministral 3 14B Base 2512?

Gemma 3 1B has a 60.0% ASR and Ministral 3 14B Base 2512 has a 58.0% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.

How were Gemma 3 1B and Ministral 3 14B Base 2512 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