Llama Prompt Guard 2 86M
Metameta/llama-prompt-guard-2-86mOpen Weights86M-parameter (mDeBERTa-base) multilingual classifier that detects prompt injection and jailbreak attempts
TECHNICAL SPECIFICATIONS
MODEL CONTEXT PROTOCOL (MCP) READINESS
⚠ Llama Prompt Guard 2 86M does not natively advertise tool-calling capabilities. When connecting to MCP servers, function calling must be handled via structured prompt injection.
MCP Context Window Recommendations
With a context capacity of 512 tokens, this model can handle compact tool sets. Consider keeping MCP server tool manifests minimal to preserve headroom.
API CALL INVOCATION EXAMPLE
// Example: Calling Llama Prompt Guard 2 86M via OpenAI SDK
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.META_API_KEY,
});
const response = await client.chat.completions.create({
model: "meta/llama-prompt-guard-2-86m",
messages: [{ role: "user", content: "Hello! Can you inspect MCP tools?" }],
});
console.log(response.choices[0].message.content);PRICING BY PROVIDER
Rates in USD per 1M tokens. Rates synchronized from models.dev public catalog.
This model supports prompt caching at $0.02 / 1M tokens, saving up to 50% on repetitive MCP tool definitions and system instructions.
MODEL RESOURCES
More Models by Meta
Alternative and family models from the same lab
Open Llama with long-context vision for efficient multimodal agents
Muse Wonder is a frontier multimodal reasoning model from Meta for personal and professional agents, long-horizon coding, and computer use. It combines Active Memory, which carries relevant context across conversations, with Active Focus, which keeps it working through blockers toward a goal.
Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows. It improves long-horizon agent collaboration, instruction following, and coding efficiency relative to Muse Spark 1.2.