Decision models · Meta

Llama Prompt Guard 2 86M

Name: Llama Prompt Guard 2 86M
Author: Meta

Task: Safety and policy checks
Returns: Probability or score
Precision tier: 76 / 100
Size: 280M
Max input: 512 tokens
License: llama4
Released: 2025-04

What it decides

Returns: benign or malicious (prompt injection or jailbreak), with a score.
Checks: Prompts and tool outputs (injection)
Languages: 8 languages
Precision: 76 / 100 · AUC 0.998 and 97.5% recall at 1% false positives in English (model card).
Clarity: Probability or score (15 of 15 clarity points)
How it runs: Classifier head · reads the input once and writes nothing, so it answers in milliseconds.
License: llama4 (Custom (read the terms))

Source: model card on Hugging Face (opens in a new tab) · How decision models are rated

How to run it

Original weights · Hugging Face

Full-precision files from Meta, for Transformers, Sentence Transformers, Text Embeddings Inference, Infinity or vLLM.

huggingface.co/meta-llama/Llama-Prompt-Guard-2-86M (opens in a new tab)

Free Hugging Face account needed: request access on the model page, and Meta approves it (usually within minutes to a few hours).

Links checked 3 days ago against Hugging Face, Ollama and GitHub.

Apps that run Llama Prompt Guard 2 86M

Our pick for each kind of computer, assuming the model fits in its memory. Pick your own GPU or Mac in the Model Finder for an exact answer.

Mac (Apple silicon)

Infinity

Local API for rerankers and classifiers, including models TEI does not support.

Also good: Sentence Transformers

Windows PC, NVIDIA GPU

Sentence Transformers

A few lines of Python with Transformers or Sentence Transformers; easiest for scripts and notebooks.

Also good: Infinity

Windows PC, AMD Radeon

Sentence Transformers

A few lines of Python with Transformers or Sentence Transformers; easiest for scripts and notebooks.

Also good: Infinity

Linux, NVIDIA GPU

Infinity

Local API for rerankers and classifiers, including models TEI does not support.

Also good: Sentence Transformers

Linux server, 80 GB+ GPUs

Infinity

Local API for rerankers and classifiers, including models TEI does not support.

Also good: Sentence Transformers

App Type Opens it as Systems Engine Start

Infinity

Server

Original weights

WSL

PyTorch, ONNX, CTranslate2

infinity_emb v2 --model-id meta-llama/Llama-Prompt-Guard-2-86M

Sentence Transformers

Library

Original weights

PyTorch

python -c "from transformers import pipeline; print(pipeline('text-classification', model='meta-llama/Llama-Prompt-Guard-2-86M')('text'))"

Which apps can open a model depends on the files published for it: GGUF for llama.cpp-based apps, MLX for Apple’s engine, the original weights for vLLM and SGLang. Compare every app.

Memory and speed

Each precision on an NVIDIA GeForce RTX 4090, for one decision on a 512-token input. The best hardware table below covers other GPUs and Macs.

Format	Memory needed	Time per decision	Decisions per second	Quality	Download
`Q8_0`	0.8 GB	6.6 ms	151/s	76Great quality (A) · 76 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`F16`	1.0 GB	7.0 ms	143/s	76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	Original weights (opens in a new tab)

One request at a time. Servers that batch (Text Embeddings Inference, vLLM, Infinity) handle many more decisions per second.

Best hardware for Llama Prompt Guard 2 86M

The least expensive GPU in our catalog that runs it without offloading is the Intel Arc A380 (launch price $139) at F16.

Quality	Hardware	Quant	Memory	Time per decision	Runs
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	AMD Radeon RX 7900 XTX	`F16`	1.2 GB / 24.0 GB	6.4 ms	Runs great
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	AMD Radeon RX 9070 XT	`F16`	1.2 GB / 16.0 GB	7.0 ms	Runs great
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	AMD Ryzen AI Max+ 395 (128 GB)	`F16`	1.2 GB / 96.0 GB	8.9 ms	Runs great
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	Apple M1 Max (64 GB)	`F16`	1.2 GB / 48.0 GB	32 ms	Runs great
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	Apple M2 Ultra (192 GB)	`F16`	1.2 GB / 144 GB	15 ms	Runs great
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	Apple M3 Ultra (256 GB)	`F16`	1.2 GB / 192 GB	14 ms	Runs great
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	Apple M3 Ultra (512 GB)	`F16`	1.2 GB / 384 GB	14 ms	Runs great
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	Apple M4 (16 GB)	`F16`	1.2 GB / 10.7 GB	71 ms	Runs great
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	Apple M4 (24 GB)	`F16`	1.2 GB / 16.1 GB	71 ms	Runs great
76Great quality (A) · 76 of 100: how capable the model is at F16, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	Apple M4 Max (128 GB)	`F16`	1.2 GB / 96.0 GB	20 ms	Runs great

Check it on your own hardware

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.