GlubBot 1.1 (7.5M)

GlubBot 1.1 is a small language model I trained mostly to test my training pipeline and see what I could get out of a tiny model. It’s only 7.5 million parameters and was trained on the TinyStories dataset, so don’t expect it to compete with larger models.

The goal wasn’t to build the smartest chatbot—it was to learn more about training language models and get something working in Ollama.

Running

ollama run caplette776/glubbot1.1-7.5m

What it’s good for

• Testing
• Playing around with a tiny LLM
• Learning how language models work
• Seeing what a model this small can do

What it’s not good at

• Long conversations
• Complex reasoning
• Coding
• Reliable factual answers

Since it’s trained on TinyStories, it tends to do best with simple writing and short responses.

Why I made it

This was mainly a test project. I wanted to experiment with training a small language model from scratch and publishing it through Ollama. More capable versions may come later as I continue experimenting.

Created by Landon Caplette