Model Card for Model ID

Model Details

Model Description

Small LLM model finetuned for a Cuse Cyber Games challenge.

Uses

Designed for the Spring 2026 Cuse Cyber Games, hosted by the Information Security Club at Syracuse University

It runs on an M1 Macbook Air 8gb wt 80 tok/s, so should have no problem running on minimal hardware.

Recommendations

Reduce the max output length - it tends to loop.

These settings are best: num_predict: 256 num_keep: 1

Set those parameters after running the model with /set parameter parameter_name value

How to Get Started with the Model

Install ollama Run the model Get the flag

Training Details

Finetuned from the base model SmolLM2-135M using open-source data about pirates from Project Gutenburg and Wikipedia.

Downloads last month
146
GGUF
Model size
0.1B params
Architecture
llama
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for TownsendSP/smolLM-135M-piratecusectf

Quantized
(41)
this model