gemmaiku-3-1b-it-GGUF-experimental

Gemmaiku 1B IT - GGUF

A small model speaks,
Five, seven, and five again,
Haiku form is met.

This repository contains the GGUF format weights for Gemmaiku-3-1b-it, a fine-tuned version of Google's Gemma-3-1b-it specialized in generating traditional 5-7-5 syllable haikus on any topic.

The SFT / MLX model weights can be found in the main repository:

vi-c0de/gemmaiku-3-1b-it-experimental.

Files Available

• gemmaiku-1b.gguf: F16 precision weights. Ideal for full-quality local evaluation on Apple Silicon, Windows, and Linux.

---

Running the Model

1. Via Ollama (Direct Load)

Ollama allows you to run GGUF models directly from Hugging Face without manual setup! Run:

ollama run hf.co/vi-c0de/gemmaiku-3-1b-it-GGUF-experimental

2. Via Custom Ollama Modelfile

If you want to download the gemmaiku-1b.gguf file locally and load it into Ollama manually with custom system instructions, create a Modelfile in the folder where your GGUF file is located:

FROM ./gemmaiku-1b.gguf

SYSTEM "You are a specialized Haiku bot. You only speak in 3 lines of 5-7-5 syllables. No chatter."

TEMPLATE """<start_of_turn>user
{{ .Prompt }}<end_of_turn>
<start_of_turn>model
"""

PARAMETER stop "<end_of_turn>"
PARAMETER temperature 0.3
PARAMETER repeat_penalty 1.1
PARAMETER top_p 0.9

Then build and run the model:

ollama create gemmaiku -f Modelfile
ollama run gemmaiku