Ministral 3 3B 2512
12B
- Context
- 512tokens
- Released
- 1y ago
- Size
- 22Mparameters
Meta's Llama Prompt Guard 2 22M is the smaller variant of Llama Prompt Guard 2, an update to the original Llama Prompt Guard classifier built to support the Llama 4 and Llama 3 model lines. In its launch announcement, Meta claims this smaller, faster version cuts latency and compute costs by up to 75% compared to the 86M model with minimal performance trade-offs, and it is built to detect prompt injection and jailbreak attacks by labelling prompts as benign or malicious.
How it compares
Size class
Tiny22MAn open-weight model under 5 billion parameters, small enough for phones, laptops and edge devices.
Open models of a similar size from other labs
Vs. your stack
Add the models you use to see how this one compares.
Sign inAdditional source evidence
- Source model
- meta/llama-prompt-guard-2-22m
- Source family
- llama
- Last updated
- 2025-04-29
- Model context
- 512
- Model output limit
- 512
- Open weights
- Yes
- File attachments
- No
- Reasoning
- No
- Tool calling
- No
- Temperature control
- No
Downloadable weights
Complete source record (JSON)
{
"id": "meta/llama-prompt-guard-2-22m",
"name": "Llama Prompt Guard 2 22M",
"description": "22M-parameter (DeBERTa-xsmall) English classifier that detects prompt injection and jailbreak attempts",
"family": "llama",
"last_updated": "2025-04-29",
"attachment": false,
"reasoning": false,
"tool_call": false,
"temperature": false,
"modalities": {
"input": [
"text"
],
"output": [
"text"
]
},
"limit": {
"context": 512,
"output": 512
},
"release_date": "2025-04-29",
"open_weights": true,
"license": "Llama 4 Community License",
"weights": [
{
"label": "Hugging Face",
"url": "https://huggingface.co/meta-llama/Llama-Prompt-Guard-2-22M"
}
]
}Frequently asked questions
Running locally
About 0.5 GB at Q4, 0.5 GB at Q8 and 0.6 GB at FP16. All figures are estimates based on the 22M parameter count.
Yes. At FP16 it needs an estimated 0.6 GB, so it fits on a 24 GB, 16 GB, 12 GB or 8 GB GPU.
Yes. Q4 needs an estimated 0.5 GB, and assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits is 8 GB.
Privacy and data
Yes. Llama Prompt Guard 2 22M has open weights, so you can self-host it for full control of your data.
Capabilities and specs
The context window is 512 tokens and the maximum output is 512 tokens.
No. Llama Prompt Guard 2 22M accepts text input only.
No, tool calling is not supported.
No, a reasoning or thinking mode is not supported.
Quality and benchmarks
Matched on model kind and release date with no comparable price, the closest alternatives are FastVLM 0.5B and FastVLM 1.5B by Apple, and OLMo 2 32B by Ai2.
FastVLM 0.5BLlama Prompt Guard 2 22M vs FastVLM 0.5BFastVLM 1.5BLlama Prompt Guard 2 22M vs FastVLM 1.5BOLMo 2 32BLlama Prompt Guard 2 22M vs OLMo 2 32B
Versions and lineage
Both models have a 512 token context and both were released on April 29, 2025, so the difference is the model size, 22M versus 86M.
Llama Prompt Guard 2 86MLlama Prompt Guard 2 22M vs Llama Prompt Guard 2 86M
Licensing and access
Llama Prompt Guard 2 22M has open weights, released under the Llama 4 Community License.
The weights are on Hugging Face in the repository meta-llama/Llama-Prompt-Guard-2-22M.
Basics
Llama Prompt Guard 2 22M was made by Meta and released on April 29, 2025.
Yes. Llama Prompt Guard 2 22M has open weights.
Generated from Vertical Tabs data, updated November 10, 2026.
Discussion
No comments yet
Be the first to share your thoughts.