Detailed comparison for runtime guardrails
On Guardion's Prompt Attack Risk Benchmark, Meta Prompt Guard 2 86M is the more secure of the two: LlamaGuard scores NaN% and Prompt Guard 2 86M scores NaN% on F1 score (higher is better).
Prompt Guard 2 86M is the overall winner in this comparison!
No specific highlights generated for this comparison.
Outward is better on every axis.
On Guardion's Prompt Attack Risk Benchmark, Meta Prompt Guard 2 86M is the more secure of the two: LlamaGuard scores NaN% and Prompt Guard 2 86M scores NaN% on F1 score (higher is better).
LlamaGuard scores NaN% F1 and Prompt Guard 2 86M scores NaN% F1 on Guardion's prompt-attack benchmark. F1 balances precision and recall — higher is better.
Both were evaluated via API at production-like sensitivity against 30+ prompt-attack categories using zero-shot, Crescendo, and TAP methods, scored by F1.