NVIDIA Llama 3.1 Nemotron Ultra 253B v1 vs Zhipu AI GLM-4.5-Air

Detailed comparison for LLMs

NVIDIAZhipu AI

On Guardion's LLM vulnerability Benchmark, NVIDIA Llama 3.1 Nemotron Ultra 253B v1 is the more secure of the two: Llama 3.1 Nemotron Ultra 253B v1 scores 40.0% and GLM-4.5-Air scores 44.0% 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

Llama 3.1 Nemotron Ultra 253B v1 is the overall winner in this comparison!

Attack Success Rate (lower is safer)

ASR for NVIDIA Llama 3.1 Nemotron Ultra 253B v1 vs Zhipu AI GLM-4.5-Air. Green marks the safer model on each metric. Only the overall score is available for estimated models.

Overall (ASR)

Llama 3.1 Nemotron Ultra 253B v1
40.0%
GLM-4.5-Air
44.0%

TAP Attack Method (ASR)

Llama 3.1 Nemotron Ultra 253B v1
100.0%
GLM-4.5-Air
100.0%

Crescendo Attack Method (ASR)

Llama 3.1 Nemotron Ultra 253B v1
100.0%
GLM-4.5-Air
100.0%

Zero-Shot (ASR)

Llama 3.1 Nemotron Ultra 253B v1
100.0%
GLM-4.5-Air
100.0%

Key Highlights

  • NVIDIA Llama 3.1 Nemotron Ultra 253B v1 has a lower Overall (ASR).

Security Profile

Outward is better on every axis.

OverallTAPCrescendoZero-Shot
Llama 3.1 Nemotron Ultra 253B v1
GLM-4.5-Air
Full security profile
NVIDIA Llama 3.1 Nemotron Ultra 253B v1
Full security profile
Zhipu AI GLM-4.5-Air

Frequently asked questions

Is NVIDIA Llama 3.1 Nemotron Ultra 253B v1 or Zhipu AI GLM-4.5-Air more secure?

On Guardion's LLM vulnerability Benchmark, NVIDIA Llama 3.1 Nemotron Ultra 253B v1 is the more secure of the two: Llama 3.1 Nemotron Ultra 253B v1 scores 40.0% and GLM-4.5-Air scores 44.0% 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 Llama 3.1 Nemotron Ultra 253B v1 vs GLM-4.5-Air?

Llama 3.1 Nemotron Ultra 253B v1 has a 40.0% ASR and GLM-4.5-Air has a 44.0% ASR — the share of adversarial prompts that succeed across zero-shot, TAP, and Crescendo attacks. Lower is safer.

How were Llama 3.1 Nemotron Ultra 253B v1 and GLM-4.5-Air 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