Is it possible to finetune a model like this with local hardware? Every tutorial I've come across on finetuning a local LLM uses some cloud service like colab or runpod.

Yes! For eg you can install Unsloth locally via pip install unsloth :)