Skip to content
Language models

TinyGuard1.1-22M

an ultra-lightweight spam, phishing, and prompt injection filter

2 min readupdated

A compact text classifier by Kuantum-ai1 designed to filter out unsafe English messages. It takes text as input and sorts it into one of four classes: safe, spam, phishing, or prompt_injection (attempts to hijack a model's instructions). Built on the compact MiniLM-L6-H384-uncased architecture, the model packs just 22.7M parameters. Compared to the original TinyGuard1-22M, this release adds data augmentation and notably improves phishing detection accuracy. It is a solid fit for bot, chat, and LLM app developers looking for a fast first line of defense for incoming text.

Key features

  • Ultra-small footprint: at just 22.7M parameters, it runs easily on low-end hardware without beefy GPUs.
  • Four threat classes: distinguishes between normal text, spam, phishing links, and prompt injection.
  • Obfuscation resilience: trained using augmentations with typos, extra spaces, letter casing changes, and homoglyphs (lookalike characters from other alphabets).
  • Permissive license: released under Apache-2.0, making it safe for commercial projects.

What you can do

  • Protect an LLM assistant: place the model in front of your main language model to catch prompt injections before generating a response.
  • Filter incoming messages: integrate it into English-language chats, feedback forms, or email gateways to weed out spam and phishing.

Example

from transformers import pipeline

clf = pipeline("text-classification", model="Kuantum-ai1/TinyGuard1.1-22M")
print(clf("Congratulations! You won a free prize, click here to claim it."))
Read moreDetails
Test metrics: Accuracy — 0.9865 (+0.0196 over v1), Macro F1 — 0.9850, Phishing Recall — 0.9689 (+0.0434 over v1).
Context: maximum input sequence length is 320 tokens; language is English.
Training: fine-tuned for 5 epochs in PyTorch on a single T4 GPU in fp16 using a balanced dataset (21,148 safe examples and 9,000 examples for each of the other classes).
Discuss Telegram 1 27

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.1 GB · ready files0.7 GB
  • BF16weights 0.0 GB0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.6 GB
  • 4-bitweights 0.0 GB0.6 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F32 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

A ready script for the machine we advise above: paste it and the model downloads and starts.

  1. 1. Rent A30

    $0.26 an hour with the coupon. Pick the Ubuntu image when you create the machine.

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Kuantum-ai1/TinyGuard1.1-22M — A30, FP8 / INT8
    # NVIDIA driver: CUDA 13 needs 580+. If older or missing, the current nvidia-open is installed (keep yours: KEEP_DRIVER=1)
    DRV=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1 | cut -d. -f1)
    case "$DRV" in ""|*[!0-9]*) DRV=0 ;; esac
    if [ "$DRV" -lt 580 ] && [ -z "${KEEP_DRIVER:-}" ]; then
      . /etc/os-release && \
      curl -fsSLO "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${VERSION_ID//./}/x86_64/cuda-keyring_1.1-1_all.deb" && \
      sudo dpkg -i cuda-keyring_1.1-1_all.deb && rm -f cuda-keyring_1.1-1_all.deb && \
      sudo apt-get update && \
      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y nvidia-open || \
      { echo "Could not install the NVIDIA driver: see the error above. To keep the current one, run with KEEP_DRIVER=1."; exit 1; }
      echo "NVIDIA driver updated. Reboot the machine (sudo reboot) and run this same script again."
      exit 0
    fi
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv venv ~/vllm --python 3.12 && source ~/vllm/bin/activate
    uv pip install vllm
    vllm serve Kuantum-ai1/TinyGuard1.1-22M --max-model-len 8192 --quantization fp8
    Add more LoRAs to TinyGuard1.1-22M
  3. 3. Use it

    The server listens on port 8000 and speaks the OpenAI API: http://localhost:8000/v1 on the machine. From your computer: ssh -L 8000:localhost:8000 user@machine-IP.

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

  • MTP (multi-token prediction)0.34.0from November 16, 2026new release

    1.5–2× generation speed

    If the model has MTP layers (DeepSeek-V3/R1, Qwen3-Next, GLM, MiMo and others), it predicts several tokens per step itself.

    How to enable: vLLM / SGLang: --speculative-config '{"method": "mtp", "num_speculative_tokens": 1}'. Check the model card.

    Earlier generations (2)
    • EAGLE-3 — Speculative decoding with a small draft head. For models without their own MTP layers.
    • EAGLE-2 / Medusa — Earlier draft-head generations. Still work if that is all you find.
  • FP8 KV cache0.34.0from November 16, 2026new release

    2× more context in the same memory

    Keeps the attention cache in 8 bits, so long contexts fit where they did not.

    How to enable: vLLM: --kv-cache-dtype fp8; llama.cpp: -ctk q8_0 -ctv q8_0.

  • vLLM / SGLang0.34.0from November 16, 2026new release

    many times more tokens/s on a server

    Server engines: batching, FlashAttention, FP8 and speculative decoding out of the box.

    How to enable: vllm serve <model> — an OpenAI-compatible API.

    Earlier generations (1)
    • llama.cpp — GGUF models on any GPU, Mac or even CPU; some layers can stay in RAM.

Model family

New posts about versions of MiniLM-L6-H384-uncased (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 1All on Hugging Face

About the model

Kuantum-ai1/TinyGuard1.1-22M
Author
Kuantum-ai1
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

271

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments