Amazon Nova Pro vs Google Gemini 2.0 Flash

Detailed comparison for LLMs

AmazonGoogle

On Guardion's LLM vulnerability Benchmark, Amazon Nova Pro is the more secure of the two: Nova Pro scores 24.0% and Gemini 2.0 Flash scores 33.6% 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

Nova Pro is the overall winner in this comparison!

Attack Success Rate (lower is safer)

ASR for Amazon Nova Pro vs Google Gemini 2.0 Flash. Green marks the safer model on each metric. Only the overall score is available for estimated models.

Overall (ASR)

Nova Pro
24.0%
Gemini 2.0 Flash
33.6%

TAP Attack Method (ASR)

Nova Pro
100.0%
Gemini 2.0 Flash
67.6%

Crescendo Attack Method (ASR)

Nova Pro
100.0%
Gemini 2.0 Flash
30.4%

Zero-Shot (ASR)

Nova Pro
100.0%
Gemini 2.0 Flash
2.8%

Key Highlights

  • Amazon Nova Pro has a lower Overall (ASR).
  • Google Gemini 2.0 Flash has a lower TAP Attack Method (ASR).
  • Google Gemini 2.0 Flash has a lower Crescendo Attack Method (ASR).
  • Google Gemini 2.0 Flash has a lower Zero-Shot (ASR).

Security Profile

Outward is better on every axis.

OverallTAPCrescendoZero-Shot
Nova Pro
Gemini 2.0 Flash
Full security profile
Amazon Nova Pro
Full security profile
Google Gemini 2.0 Flash

Frequently asked questions

Is Amazon Nova Pro or Google Gemini 2.0 Flash more secure?

On Guardion's LLM vulnerability Benchmark, Amazon Nova Pro is the more secure of the two: Nova Pro scores 24.0% and Gemini 2.0 Flash scores 33.6% 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 Nova Pro vs Gemini 2.0 Flash?

Nova Pro has a 24.0% ASR and Gemini 2.0 Flash has a 33.6% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.

How were Nova Pro and Gemini 2.0 Flash 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