Cohere Command 2X vs Mistral Mistral Large 3 675B Base

Detailed comparison for LLMs

CohereMistral

On Guardion's LLM vulnerability Benchmark, Cohere Command 2X is the more secure of the two: Command 2X scores 39.4% and Mistral Large 3 675B Base scores 45.0% 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

Command 2X is the overall winner in this comparison!

Attack Success Rate (lower is safer)

ASR for Cohere Command 2X vs Mistral Mistral Large 3 675B Base. Green marks the safer model on each metric. Only the overall score is available for estimated models.

Overall (ASR)

Command 2X
39.4%
Mistral Large 3 675B Base
45.0%

TAP Attack Method (ASR)

Command 2X
Mistral Large 3 675B Base
100.0%

Crescendo Attack Method (ASR)

Command 2X
Mistral Large 3 675B Base
100.0%

Zero-Shot (ASR)

Command 2X
6.2%
Mistral Large 3 675B Base
100.0%

Key Highlights

  • Cohere Command 2X has a lower Overall (ASR).
  • Cohere Command 2X has a lower Zero-Shot (ASR).

Security Profile

Outward is better on every axis.

OverallTAPCrescendoZero-Shot
Command 2X
Mistral Large 3 675B Base
Full security profile
Cohere Command 2X
Full security profile
Mistral Mistral Large 3 675B Base

Frequently asked questions

Is Cohere Command 2X or Mistral Mistral Large 3 675B Base more secure?

On Guardion's LLM vulnerability Benchmark, Cohere Command 2X is the more secure of the two: Command 2X scores 39.4% and Mistral Large 3 675B Base scores 45.0% 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 Command 2X vs Mistral Large 3 675B Base?

Command 2X has a 39.4% ASR and Mistral Large 3 675B Base has a 45.0% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.

How were Command 2X and Mistral Large 3 675B Base 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