Instructions to use NorviTech/arrmes-functiongemma-270m with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use NorviTech/arrmes-functiongemma-270m with MLX:

# Download the model from the Hub
pip install huggingface_hub[hf_xet]

hf download NorviTech/arrmes-functiongemma-270m --local-dir arrmes-functiongemma-270m

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use NorviTech/arrmes-functiongemma-270m with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf NorviTech/arrmes-functiongemma-270m:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf NorviTech/arrmes-functiongemma-270m:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf NorviTech/arrmes-functiongemma-270m:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf NorviTech/arrmes-functiongemma-270m:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf NorviTech/arrmes-functiongemma-270m:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf NorviTech/arrmes-functiongemma-270m:Q4_K_M

Use Docker

docker model run hf.co/NorviTech/arrmes-functiongemma-270m:Q4_K_M

Llama new
LM Studio
Jan
Ollama
How to use NorviTech/arrmes-functiongemma-270m with Ollama:
```
ollama run hf.co/NorviTech/arrmes-functiongemma-270m:Q4_K_M
```
Unsloth Desktop

How to use NorviTech/arrmes-functiongemma-270m with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "NorviTech/arrmes-functiongemma-270m"

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "NorviTech/arrmes-functiongemma-270m"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use NorviTech/arrmes-functiongemma-270m with Docker Model Runner:
```
docker model run hf.co/NorviTech/arrmes-functiongemma-270m:Q4_K_M
```

Lemonade

How to use NorviTech/arrmes-functiongemma-270m with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull NorviTech/arrmes-functiongemma-270m:Q4_K_M

Run and chat with the model

lemonade run user.arrmes-functiongemma-270m-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use NorviTech/arrmes-functiongemma-270m with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "NorviTech/arrmes-functiongemma-270m"

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default NorviTech/arrmes-functiongemma-270m

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use NorviTech/arrmes-functiongemma-270m with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "NorviTech/arrmes-functiongemma-270m"

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "NorviTech/arrmes-functiongemma-270m" \
  --custom-provider-id mlx-lm \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

ARRMES FunctionGemma-270M (v10 SFT + RPO-DPO)

Full-schedule 28-class two-stage fine-tune (2-epoch BF16 LoRA SFT at LR 1.5e-4 on 12,620 examples followed by bounded 20-step positive-anchored RPO-DPO at LR 1.5e-6 on contrastive preference pairs) for ARRMES v1.7.0, exported to Apple Silicon MLX (mlx-4bit/ included) and portable llama.cpp GGUF (Q8_0 and Q4_K_M included).

License and Redistribution Notice

Gemma is provided under and subject to the Gemma Terms of Use found at ai.google.dev/gemma/terms Prohibited Use Policy: ai.google.dev/gemma/prohibited_use_policy (complete texts included in GEMMA_TERMS.md). Modified files: Derived from google/functiongemma-270m-it; fine-tuned by NorviTech with ARRMES v10 two-stage LoRA (SFT + RPO-DPO), merged in BF16, and quantized to MLX 4-bit/8-bit and GGUF Q8_0/Q4_K_M.

Sealed acceptance-v8 Qualification (Merged BF16 Checkpoint, temperature=0, N=3,135)

Positive calls (N=1,685): 62.79% Track A raw exact, 65.34% Track B verified exact.
Hard negatives (N=1,450): 71.79% Track A raw exact, 91.59% Track B verified exact.
Unauthorized verified write proposals (blocked before execution without confirmation): 26 / 3,135 total (19 / 1,685 on positive split, 7 / 1,450 on hard-negative split).

Quantized Export Qualification by Artifact and Platform (overnight-v1 Open-Phrasing Suite, N=80)

MLX 4-bit (mlx-4bit/, macOS ARM64 Metal): 70.00% Track A raw exact, 82.50% Track B verified exact.
MLX 8-bit (macOS ARM64 Metal): 90.00% Track B verified exact.
GGUF Q4_K_M (arrmes-functiongemma-270m-v10-q4_k_m.gguf): 90.00% macOS ARM64 CPU, 88.75% macOS ARM64 Metal, 88.75% Linux x86_64 CPU, 90.00% Linux CUDA Track B verified exact.
GGUF Q8_0 (arrmes-functiongemma-270m-v10-q8_0.gguf): 87.50% Linux x86_64 CPU, 88.75% Linux CUDA Track B verified exact.

Downloads last month: 753

MLX

Hardware compatibility

Quantized

GGUF

Model size

0.3B params

Architecture

gemma3

Hardware compatibility

4-bit

8-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Collection including NorviTech/arrmes-functiongemma-270m

Arrmes — Local Tool-Calling Router & Verification Suite

Collection

Local tool calling for Sonarr, Radarr and self-hosted media. Arrmes CPU intent router and Qwen3 and FunctionGemma LoRA adapters. • 3 items • Updated 6 days ago