Instructions to use Open4bits/gemma-3-270m-it-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Open4bits/gemma-3-270m-it-gguf with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Open4bits/gemma-3-270m-it-gguf")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Open4bits/gemma-3-270m-it-gguf", dtype="auto")

llama-cpp-python

How to use Open4bits/gemma-3-270m-it-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Open4bits/gemma-3-270m-it-gguf",
	filename="gemma-3-270m-it-IQ3_M-imat.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Open4bits/gemma-3-270m-it-gguf with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Open4bits/gemma-3-270m-it-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Open4bits/gemma-3-270m-it-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Open4bits/gemma-3-270m-it-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Open4bits/gemma-3-270m-it-gguf:Q4_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 Open4bits/gemma-3-270m-it-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Open4bits/gemma-3-270m-it-gguf:Q4_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 Open4bits/gemma-3-270m-it-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Open4bits/gemma-3-270m-it-gguf:Q4_K_M

Use Docker

docker model run hf.co/Open4bits/gemma-3-270m-it-gguf:Q4_K_M

LM Studio
Jan

vLLM

How to use Open4bits/gemma-3-270m-it-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Open4bits/gemma-3-270m-it-gguf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Open4bits/gemma-3-270m-it-gguf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Open4bits/gemma-3-270m-it-gguf:Q4_K_M

SGLang

How to use Open4bits/gemma-3-270m-it-gguf with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Open4bits/gemma-3-270m-it-gguf" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Open4bits/gemma-3-270m-it-gguf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Open4bits/gemma-3-270m-it-gguf" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Open4bits/gemma-3-270m-it-gguf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Ollama
How to use Open4bits/gemma-3-270m-it-gguf with Ollama:
```
ollama run hf.co/Open4bits/gemma-3-270m-it-gguf:Q4_K_M
```

Unsloth Studio new

How to use Open4bits/gemma-3-270m-it-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 Open4bits/gemma-3-270m-it-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 Open4bits/gemma-3-270m-it-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Open4bits/gemma-3-270m-it-gguf to start chatting

Docker Model Runner
How to use Open4bits/gemma-3-270m-it-gguf with Docker Model Runner:
```
docker model run hf.co/Open4bits/gemma-3-270m-it-gguf:Q4_K_M
```

Lemonade

How to use Open4bits/gemma-3-270m-it-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Open4bits/gemma-3-270m-it-gguf:Q4_K_M

Run and chat with the model

lemonade run user.gemma-3-270m-it-gguf-Q4_K_M

List all available models

lemonade list

gemma-3-270m-gguf

This repository contains GGUF-formatted, quantized versions of Gemma 3 270M (base model), prepared and published by Open4Bits for efficient local inference on low-resource systems.

Open4Bits is a model distribution and optimization initiative under ArkAILabs, which is run and operated by ArkDevLabs. Through Open4Bits, ArkAI Labs publishes quantized, optimized, and deployment-ready models in GGUF and other inference-friendly formats.

About Open4Bits

Open4Bits focuses on making modern language models usable on real-world hardware.

Through Open4Bits, ArkAI Labs publishes:

Quantized language models
GGUF models for local inference
Optimized formats for CPU-friendly deployment
Lightweight variants for low-resource systems

The goal is to enable practical AI usage without requiring high-end GPUs.

Available Models

File	Quantization	Notes
`gemma-3-270m-IQ4_NL.gguf`	IQ4_NL	Ultra-light, fastest, lowest memory usage
`gemma-3-270m-IQ4_XS.gguf`	IQ4_XS	Slightly higher quality than IQ4_NL
`gemma-3-270m-Q4_0.gguf`	Q4_0	Legacy 4-bit quantization
`gemma-3-270m-Q4_1.gguf`	Q4_1	Legacy 4-bit with improved accuracy
`gemma-3-270m-Q4_K_S.gguf`	Q4_K_S	Modern K-quant (small)
`gemma-3-270m-Q4_K_M.gguf`	Q4_K_M	Modern K-quant (medium, recommended)
`gemma-3-270m-Q5_0.gguf`	Q5_0	Legacy 5-bit quantization
`gemma-3-270m-Q5_1.gguf`	Q5_1	Legacy 5-bit with improved accuracy
`gemma-3-270m-Q5_K_S.gguf`	Q5_K_S	Modern K-quant (small, higher quality)
`gemma-3-270m-Q5_K_M.gguf`	Q5_K_M	Modern K-quant (medium, high quality)
`gemma-3-270m-Q8_0.gguf`	Q8_0	Near-baseline quality (reference)

Recommended Variants

Q4_K_M – Recommended for most users (best balance of quality, speed, and memory)
Q5_K_M – Higher quality for users who can afford additional memory usage
Q8_0 – Near-baseline reference for evaluation and comparison

Quantization Overview

IQ4 Variants

Extremely small and fast
Suitable for very limited hardware
Noticeable quality reduction compared to K-quant variants

Q4_0 / Q4_1

Older quantization formats
Included for compatibility and comparison
Generally inferior to modern K-quant variants

K-Quant Variants (Q4_K_S, Q4_K_M, Q5_K_S, Q5_K_M)

Newer, higher-quality quantization methods
Better accuracy per bit
Standard choice for GGUF deployments

Q8_0

Near-baseline (FP16-like) quality
Highest memory usage among provided variants
Intended for reference, testing, and debugging

Usage with llama.cpp

./main -m gemma-3-270m-Q4_K_M.gguf \
  -p "Explain what GGUF quantization is." \
  -n 128

Usage with Ollama

Create a Modelfile:

FROM ./gemma-3-270m-Q4_K_M.gguf

Then run:

ollama create gemma-270m -f Modelfile
ollama run gemma-270m

System Requirements (Approximate)

Quantization	Memory Requirement
IQ4_NL / IQ4_XS	~300–400 MB
Q4_0 / Q4_1	~450–500 MB
Q4_K_S	~500 MB
Q4_K_M	~550–600 MB
Q5_K_S	~600–650 MB
Q5_K_M	~650–750 MB
Q8_0	~900 MB – 1 GB

Actual memory usage depends on context length and runtime configuration.

Intended Use and Limitations

This model is a small, lightweight base language model optimized for low-resource inference.

It is suitable for:

Token generation and experimentation
Embedding-style or downstream fine-tuning use
Educational and demonstration purposes
Lightweight local inference

Limitations:

Not instruction-tuned
Not designed for complex reasoning tasks
Not suitable for advanced code generation
May hallucinate or produce incomplete outputs for complex prompts

License

This repository follows the original Gemma model license.

Users are responsible for complying with all upstream license terms and usage restrictions.

Credits

Original model: Google (Gemma)
Quantization and GGUF packaging: Open4Bits (ArkAI Labs, operated by ArkDevLabs)
Tooling: llama.cpp

Support

Open4Bits is an initiative under ArkAI Labs, operated by ArkDevLabs, focused on open and accessible AI.

If you find these models useful, please consider supporting the work by:

Liking the model on Hugging Face
Sharing feedback or reporting issues
Using Open4Bits models in your projects and research

Organization Links

Open4Bits on Hugging Face: https://huggingface.co/Open4Bits
ArkAI Labs on Hugging Face: https://huggingface.co/ArkAiLab-Adl
ArkDevLabs Website: https://arkdevlabs.com

Downloads last month: 279

GGUF

Model size

0.3B params

Architecture

gemma3

Hardware compatibility

3-bit

4-bit

5-bit

6-bit

8-bit

16-bit

32-bit

Model tree for Open4bits/gemma-3-270m-it-gguf

Base model

google/gemma-3-270m

Finetuned

google/gemma-3-270m-it

Quantized

(89)

this model

Collection including Open4bits/gemma-3-270m-it-gguf

GGUF

Collection

28 items • Updated 19 days ago • 1