Llama Prompt Guard 2 86M

unknown

Meta · llama family

86M-parameter (mDeBERTa-base) multilingual classifier that detects prompt injection and jailbreak attempts

Limits

Context512
InputUnknown
Output512

Modalities

Inputtext
Outputtext

Release info

Release2025-04-29
Last update2025-04-29
Knowledge cutoffUnknown

Provider offerings

Provider-specific prices and limits stay separate from canonical model facts.

ProviderProvider model IDInputOutputContextOutput limit
No visible provider offerings yet.

Capabilities

Attachments

No

Reasoning

No

Tool calling

No

Structured output

Unknown

Temperature

No