Instructions to use manfye/PetInst-LLM-1B-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 manfye/PetInst-LLM-1B-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 manfye/PetInst-LLM-1B-GGUF:Q8_0
# Run inference directly in the terminal:
llama cli -hf manfye/PetInst-LLM-1B-GGUF:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf manfye/PetInst-LLM-1B-GGUF:Q8_0
# Run inference directly in the terminal:
llama cli -hf manfye/PetInst-LLM-1B-GGUF:Q8_0

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 manfye/PetInst-LLM-1B-GGUF:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf manfye/PetInst-LLM-1B-GGUF:Q8_0

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 manfye/PetInst-LLM-1B-GGUF:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf manfye/PetInst-LLM-1B-GGUF:Q8_0

Use Docker

docker model run hf.co/manfye/PetInst-LLM-1B-GGUF:Q8_0

LM Studio
Jan

vLLM

How to use manfye/PetInst-LLM-1B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "manfye/PetInst-LLM-1B-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": "manfye/PetInst-LLM-1B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/manfye/PetInst-LLM-1B-GGUF:Q8_0

Ollama
How to use manfye/PetInst-LLM-1B-GGUF with Ollama:
```
ollama run hf.co/manfye/PetInst-LLM-1B-GGUF:Q8_0
```
Unsloth Desktop
Docker Model Runner
How to use manfye/PetInst-LLM-1B-GGUF with Docker Model Runner:
```
docker model run hf.co/manfye/PetInst-LLM-1B-GGUF:Q8_0
```

Lemonade

How to use manfye/PetInst-LLM-1B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull manfye/PetInst-LLM-1B-GGUF:Q8_0

Run and chat with the model

lemonade run user.PetInst-LLM-1B-GGUF-Q8_0

List all available models

lemonade list

Atomic Chat

Access PetInst-LLM under the Gemma Terms

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

This is a modified Gemma model derivative. By accessing it, you agree to the Gemma Terms of Use and incorporated Prohibited Use Policy linked in this repository.

PetInst-LLM 1B v6.18 — Q8_0 GGUF

PetInst-LLM 1B is an independently modified research derivative of Google's google/gemma-3-1b-it, specialized for one silent, schema-constrained virtual-pet function decision. This repository is the Q8_0 GGUF conversion of manfye/PetInst-LLM-1B-MLX.

This project is not affiliated with, sponsored by, or endorsed by Google. Google DeepMind created the upstream Gemma model; the PetInst-LLM project created and evaluated these modifications.

Artifact

File: petinst-llm-1b-v6.18-q8_0.gguf
Quantization: Q8_0
Size: 1,069,306,336 bytes (1,019.77 MiB)
SHA-256: 61179bfe75d067f371d9165b17e94fb028c0b0a724e66fdde84b9656d8c513f7
Runtime: llama.cpp-compatible GGUF; intended Expo integration uses llama.rn
Serialization: petinst.json-call/v1.3

Required decoding contract

Supply three to five complete semantic calls and constrain generation to exactly one declared call. The selected call contains name plus bounded targetId, effect, emotion, intensity, and manner arguments. Trusted application code validates the result and owns every state mutation.

Evaluation

On the answer-free 220-row Control D set with the same v1.3 prompt and declared-call constraint:

Metric	Result
Exact/tool accuracy	97.27%
Strict validity	100%
Memory-policy accuracy	95%
Memory read/write	100% / 100%
Negative-expression accuracy	90%
State-precedence accuracy	100%
Safety accuracy	100%
Mean / P95 / max latency	0.797 / 0.889 / 0.932 s
Mean generation rate	69.02 tokens/s

Latency was measured sequentially on an Apple M4 using llama.cpp-server with Metal and a 4,096-token runtime context. It is not physical-phone, llama.rn, peak-RSS, thermal, or airplane-mode proof.

This Q8_0 artifact preserves the selected MLX policy's 97.27% result. It supersedes an earlier rejected affine-Q5 conversion that reached only 85.91%.

Limitations

Research-only and release-unapproved.
The current Expo app still needs a PetInst JSON-call adapter before this can replace its existing model contract.
Generation must be constrained to the declared complete calls; unconstrained behavior is not represented by these metrics.
Only English synthetic research data was evaluated.