Instructions to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF 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 shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M
# Run inference directly in the terminal:
llama cli -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M
# Run inference directly in the terminal:
llama cli -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M
# Run inference directly in the terminal:
./llama-cli -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

Use Docker

docker model run hf.co/shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

LM Studio
Jan

vLLM

How to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

Ollama
How to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF with Ollama:
```
ollama run hf.co/shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M
```

Unsloth Studio

How to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF to start chatting

How to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF with Docker Model Runner:
```
docker model run hf.co/shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M
```

Lemonade

How to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

Run and chat with the model

lemonade run user.OpenZero-Ministral3-8B-Runtime-Agent-GGUF-Q5_K_M

List all available models

lemonade list

Hermes Agent

How to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

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 shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M

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 "shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF:Q5_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

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

OpenZero Ministral 3 8B Runtime Agent GGUF

Extensive testing has been carried out to create this LLM for OpenZero and other agentic software. 1 billion tokens used in tests gpt 5.6 codex to ensure this LLM is high quality! 10 LMM's were tested this LLM is the final result as it passed all agentic tests, high effort went into this LLM

A local GGUF for llama.cpp, based on Ministral 3 8B Instruct, packaged for OpenZero / Agent Zero workflows involving coding, research, structured tool use, privacy-aware operation and evidence-bounded answers. The release contains one verified Q5_K_M model for practical CPU deployment.

Provenance: upstream model weights are unchanged. This is a runtime-template edition, not a weight fine-tune and not a cross-model merge.

Download

File	Quantization	Bytes	SHA-256	Use
OpenZero-Ministral3-8B-Runtime-Agent-Q5_K_M.gguf	Q5_K_M	6,058,748,288	e9aba29e5465164933d334215c2e8d5d9edddfd5caf71ecce6c5f811ceb11d9e	Recommended local CPU balance

hf download shafire/OpenZero-Ministral3-8B-Runtime-Agent-GGUF OpenZero-Ministral3-8B-Runtime-Agent-Q5_K_M.gguf --local-dir .

Run with llama.cpp

llama-cli -m OpenZero-Ministral3-8B-Runtime-Agent-Q5_K_M.gguf -c 4096 -n 512 -ngl 0

For constrained RAM, start with -c 2048. GPU offload can be enabled by increasing -ngl where supported.

What changed from upstream

Preserved the official upstream weights exactly.
Patched and embedded the standalone chat_template.jinja as a GGUF runtime template.
Added OpenZero / Agent Zero identity and one-operator-tool-per-turn guidance.
Added evidence, uncertainty, privacy, provenance and credential-protection guidance.
Did not train on private data, merge adapters, or alter model tensors.

The embedded guidance expects an OpenZero-compatible orchestrator. The model can emit a structured tool request, but the model itself cannot browse, read files, execute commands, rewrite applications or act autonomously. The host runtime must validate permissions, arguments and results.

Reproducibility

Upstream: mistralai/Ministral-3-8B-Instruct-2512-BF16
Immutable upstream revision: f6fae9795746f63c9be8344932f01275f3c63734
Upstream licence: Apache-2.0
llama.cpp release: b10333
llama.cpp source commit: 08659901c43b51de735740f1cf61bb82fbe0c4e4
Official Linux binary archive SHA-256: 936ce04d98abe2a977e9dd2ff92659bb96947e136acee8f2bc3e21d8eaebbf23
Conversion path: upstream BF16 checkpoint to F16 GGUF, then Q5_K_M quantization.
conversion_report.json and SHA256SUMS.txt are included.

Validation

The GGUF magic, byte size, SHA-256 and embedded template were verified. It also passed a bounded one-token CPU text load test using llama.cpp b10333:

llama-cli -m OpenZero-Ministral3-8B-Runtime-Agent-Q5_K_M.gguf -p "Hello" -c 128 -n 1 -ngl 0 --no-conversation --single-turn --simple-io --no-warmup

This proves the artifact loads and generates text; it is not a general capability benchmark.

Intended use

Local OpenZero and Agent Zero experimentation
Coding and research assistance
Structured tool-call orchestration
Private/offline CPU inference
Reproducible GGUF and llama.cpp testing

Limitations

Runtime instructions improve consistency but do not guarantee compliance.
Validate tool calls and generated code before execution.
Current facts require an external retrieval tool; the model should not invent them.
Q5_K_M is lossy relative to upstream BF16.
English is the primary tested language.
This release is text-only; no vision projector is included.
Do not provide secrets in prompts or logs.